Up next in 10
Join us for the 3.5-hour Americas Livestream of Citus Con: An Event for Postgres 2023 on CSharp TV, starting on Tue Apr 18th @ 9:00am PDT. Organized by the Postgres and Citus teams at Microsoft, Citus Con: An Event for Postgres 2023 is a free and virtual developer event with 37 talks by experts in PostgreSQL and Citus, open source users, and Azure database customers. In addition to this Americas livestream—keynoted by Simon Willison—there is also an EMEA Livestream on Wed Apr 19th @ 9:00am CEST—plus 25 more on-demand talks too. Come learn what you can do with the world’s most advanced open source relational database—from the nerdy to the sublime.
Everything you need to know about Citus Con: An Event for Postgres 2023—including the detailed schedule—can be found at: https://aka.ms/cituscon
Join the conversation You can join the #cituscon channel on Discord to join the virtual hallway track and be part of the official backchannel. And if you share learnings from the 37 talks on the socials, be sure to use hashtag #CitusCon.
C# Corner - Community of Software and Data Developers
https://www.c-sharpcorner.com
Show More Show Less View Video Transcript
0:00
foreign
0:02
foreign
0:02
foreign [Music]
0:54
thank you
0:58
[Music]
2:01
thank you
2:04
thank you
2:06
thank you
2:06
thank you [Music]
2:26
[Music]
2:26
[Music] thank you
2:29
thank you
2:29
thank you [Music]
3:29
[Music]
3:29
[Music] thank you
3:32
[Music]
3:51
[Music]
3:51
[Music] thank you
3:55
[Music]
4:06
[Music]
4:06
[Music] foreign
4:09
[Music]
5:02
thank you
5:04
thank you
5:04
thank you [Music]
5:28
thank you
5:45
[Music]
6:27
[Music]
6:39
[Music]
6:39
[Music] thank you
6:41
thank you
6:41
thank you [Music]
6:51
thank you
6:53
thank you
6:53
thank you [Music]
7:13
[Music]
7:13
[Music] foreign
7:17
[Music]
8:28
foreign
8:30
foreign
8:30
foreign [Music]
9:37
thank you
9:38
thank you
9:38
thank you foreign
9:43
[Music]
10:18
foreign
10:19
foreign
10:19
foreign [Music]
10:32
[Music]
10:32
[Music] thank you
10:37
[Music]
11:45
[Music]
11:45
[Music] thank you
11:47
thank you
11:47
thank you [Music]
12:07
thank you
12:09
thank you
12:09
thank you [Music]
12:21
thank you
12:22
thank you
12:22
thank you [Music]
13:28
[Music]
13:38
thank you
13:40
thank you
13:40
thank you thank you
13:41
thank you
13:41
thank you [Music]
14:35
[Music]
14:35
[Music] thank you
14:47
thank you
14:56
[Music]
15:34
thank you
15:36
thank you
15:36
thank you [Music]
15:56
[Music]
15:56
[Music] foreign
15:59
[Music]
16:32
thank you
16:36
[Music]
17:07
[Music]
17:07
[Music] thank you
17:09
thank you
17:09
thank you foreign
17:12
foreign
17:12
foreign [Music]
17:21
[Music]
17:21
[Music] [Applause]
17:23
[Applause]
17:23
[Applause] [Music]
18:00
[Music]
18:00
[Music] thank you
18:02
thank you
18:02
thank you [Music]
18:13
[Music]
18:13
[Music] I don't know
18:14
I don't know
18:14
I don't know another
18:17
another
18:17
another [Music]
18:53
[Music]
19:04
[Music]
19:04
[Music] thank you
19:06
thank you
19:06
thank you foreign
19:10
[Music]
19:44
[Music]
19:44
[Music] foreign
19:45
foreign
19:45
foreign [Music]
20:51
[Music]
20:51
[Music] thank you
20:55
[Music]
21:13
[Music]
21:13
[Music] thank you
21:19
[Music]
21:28
foreign
21:31
foreign
21:31
foreign [Music]
22:07
[Music]
22:33
[Music]
22:33
[Music] foreign
22:36
foreign
22:36
foreign [Music]
22:54
[Music]
22:54
[Music] [Applause]
22:55
[Applause]
22:56
[Applause] [Music]
22:59
another
23:03
season
23:04
season
23:04
season [Music]
23:12
[Music]
23:25
foreign
23:28
[Music]
24:05
thank you
24:09
[Music]
24:18
[Music]
24:57
[Music]
24:57
[Music] foreign
24:58
foreign
24:58
foreign [Music]
25:52
[Music]
25:52
[Music] thank you
25:57
[Music]
26:18
[Music]
26:18
[Music] [Applause]
26:21
[Applause]
26:21
[Applause] [Music]
26:34
[Music]
26:34
[Music] thank you
26:35
thank you
26:35
thank you [Music]
27:07
[Music]
27:07
[Music] on them
27:08
on them
27:08
on them [Music]
27:39
[Music]
27:39
[Music] thank you
27:43
[Music]
27:50
[Music]
28:30
[Music]
28:30
[Music] foreign
28:32
foreign
28:32
foreign [Music]
29:03
another level
29:06
another level
29:06
another level s
29:08
s
29:08
s [Music]
29:14
[Music]
29:53
thank you
29:55
foreign
30:00
you know I love that music I have to say
30:02
you know I love that music I have to say
30:02
you know I love that music I have to say welcome everybody thank you Isaac on my
30:05
welcome everybody thank you Isaac on my
30:06
welcome everybody thank you Isaac on my team who helped select it um my name is
30:08
team who helped select it um my name is
30:08
team who helped select it um my name is Claires rodano and this is I am Robert
30:12
Claires rodano and this is I am Robert
30:12
Claires rodano and this is I am Robert Treat
30:13
Treat
30:13
Treat and I'm a studies open source champion
30:15
and I'm a studies open source champion
30:15
and I'm a studies open source champion on the postgres team at Microsoft and
30:18
on the postgres team at Microsoft and
30:18
on the postgres team at Microsoft and you Robert R uh so I'm a former speaker
30:22
you Robert R uh so I'm a former speaker
30:22
you Robert R uh so I'm a former speaker at cytuscon this year doing the
30:24
at cytuscon this year doing the
30:24
at cytuscon this year doing the co-hosting uh otherwise a long time
30:26
co-hosting uh otherwise a long time
30:26
co-hosting uh otherwise a long time postgres Community avocado
30:30
let's dive in we've got a welcome deck
30:32
let's dive in we've got a welcome deck
30:32
let's dive in we've got a welcome deck to walk you through before we go to our
30:34
to walk you through before we go to our
30:34
to walk you through before we go to our first Speaker our keynote speaker
30:37
first Speaker our keynote speaker
30:37
first Speaker our keynote speaker um so if let's see if we can take a look
30:40
um so if let's see if we can take a look
30:40
um so if let's see if we can take a look at those slides
30:41
at those slides
30:41
at those slides um I have a bunch of things to tell you
30:44
um I have a bunch of things to tell you
30:44
um I have a bunch of things to tell you about I mean the first is that this is
30:46
about I mean the first is that this is
30:46
about I mean the first is that this is the America's live stream today and uh
30:49
the America's live stream today and uh
30:49
the America's live stream today and uh we have an emea live stream coming up at
30:51
we have an emea live stream coming up at
30:51
we have an emea live stream coming up at nine o'clock Central European summertime
30:53
nine o'clock Central European summertime
30:53
nine o'clock Central European summertime on Wednesday and then there's 25 new
30:56
on Wednesday and then there's 25 new
30:56
on Wednesday and then there's 25 new on-demand talks that were pre-recorded
30:58
on-demand talks that were pre-recorded
30:58
on-demand talks that were pre-recorded in the last couple of weeks and those
31:01
in the last couple of weeks and those
31:01
in the last couple of weeks and those just dropped on YouTube and are
31:03
just dropped on YouTube and are
31:03
just dropped on YouTube and are available for your watching pleasure now
31:05
available for your watching pleasure now
31:05
available for your watching pleasure now so you can find everything you want
31:06
so you can find everything you want
31:06
so you can find everything you want about cytuscon these two live streams
31:09
about cytuscon these two live streams
31:09
about cytuscon these two live streams and On Demand at aka.mscon
31:13
and On Demand at aka.mscon
31:13
and On Demand at aka.mscon um also I'm so nervous that I jumped
31:15
um also I'm so nervous that I jumped
31:15
um also I'm so nervous that I jumped right over the the most important thing
31:17
right over the the most important thing
31:17
right over the the most important thing which is to say that I am really excited
31:19
which is to say that I am really excited
31:19
which is to say that I am really excited about all the content all the talks all
31:22
about all the content all the talks all
31:22
about all the content all the talks all the postgres learning I mean conferences
31:24
the postgres learning I mean conferences
31:24
the postgres learning I mean conferences are such a great way to help new people
31:27
are such a great way to help new people
31:27
are such a great way to help new people learn and spin up on postgres and um
31:30
learn and spin up on postgres and um
31:30
learn and spin up on postgres and um just give back to the community so
31:32
just give back to the community so
31:32
just give back to the community so there's there's so much going on with
31:34
there's there's so much going on with
31:34
there's there's so much going on with sliderscon this year uh you know be
31:36
sliderscon this year uh you know be
31:36
sliderscon this year uh you know be helpful is that we had some kind of
31:37
helpful is that we had some kind of
31:37
helpful is that we had some kind of guide or something to sort of walk us
31:40
guide or something to sort of walk us
31:40
guide or something to sort of walk us through all the all the stuff that we
31:41
through all the all the stuff that we
31:41
through all the all the stuff that we have to look at yeah yeah well in fact
31:44
have to look at yeah yeah well in fact
31:44
have to look at yeah yeah well in fact um there is a guide It's a blog post
31:46
um there is a guide It's a blog post
31:46
um there is a guide It's a blog post check that out short URL is showing on
31:49
check that out short URL is showing on
31:49
check that out short URL is showing on the screen and there are 37 talks and it
31:51
the screen and there are 37 talks and it
31:51
the screen and there are 37 talks and it categorizes them into different buckets
31:54
categorizes them into different buckets
31:54
categorizes them into different buckets so that you can figure out which talks
31:56
so that you can figure out which talks
31:56
so that you can figure out which talks are most useful to your work and the
31:58
are most useful to your work and the
31:58
are most useful to your work and the kinds of things that you want to do with
32:00
kinds of things that you want to do with
32:00
kinds of things that you want to do with postgres and or with cytus
32:03
postgres and or with cytus
32:03
postgres and or with cytus um also just need to say thank you to
32:06
um also just need to say thank you to
32:06
um also just need to say thank you to all 40 of these amazing speakers
32:09
all 40 of these amazing speakers
32:09
all 40 of these amazing speakers yeah absolutely uh speaking and
32:11
yeah absolutely uh speaking and
32:11
yeah absolutely uh speaking and conferences it's fun it's also a lot of
32:14
conferences it's fun it's also a lot of
32:14
conferences it's fun it's also a lot of work uh I know it was more work than I
32:17
work uh I know it was more work than I
32:17
work uh I know it was more work than I could possibly do a second time so I'm
32:19
could possibly do a second time so I'm
32:19
could possibly do a second time so I'm super happy uh and say thank you to all
32:21
super happy uh and say thank you to all
32:21
super happy uh and say thank you to all these folks who have helped out uh it's
32:23
these folks who have helped out uh it's
32:23
these folks who have helped out uh it's been uh nice working with the ones that
32:24
been uh nice working with the ones that
32:24
been uh nice working with the ones that I've gotten to talk to so far and I'm
32:26
I've gotten to talk to so far and I'm
32:26
I've gotten to talk to so far and I'm definitely looking forward to seeing a
32:28
definitely looking forward to seeing a
32:28
definitely looking forward to seeing a bunch of talks
32:30
bunch of talks
32:30
bunch of talks there's a Code of Conduct as you might
32:32
there's a Code of Conduct as you might
32:32
there's a Code of Conduct as you might expect even for a virtual event like
32:34
expect even for a virtual event like
32:34
expect even for a virtual event like this one
32:35
this one
32:35
this one um and it's all the things you you might
32:37
um and it's all the things you you might
32:37
um and it's all the things you you might expect be respectful be inclusive be
32:40
expect be respectful be inclusive be
32:40
expect be respectful be inclusive be friendly and welcoming
32:41
friendly and welcoming
32:41
friendly and welcoming um if you want to see the full code of
32:43
um if you want to see the full code of
32:43
um if you want to see the full code of conduct or you need to report an issue
32:45
conduct or you need to report an issue
32:45
conduct or you need to report an issue there's a URL showing on the screen AKA
32:48
there's a URL showing on the screen AKA
32:48
there's a URL showing on the screen AKA dot Ms slash cytoscon hyphen conduct you
32:50
dot Ms slash cytoscon hyphen conduct you
32:51
dot Ms slash cytoscon hyphen conduct you can also get to the code of conduct from
32:52
can also get to the code of conduct from
32:52
can also get to the code of conduct from the footer of any one of the cytuscon
32:54
the footer of any one of the cytuscon
32:54
the footer of any one of the cytuscon web pages online
32:57
web pages online
32:57
web pages online um they're if you're watching this on
33:00
um they're if you're watching this on
33:00
um they're if you're watching this on YouTube there are live captions in
33:02
YouTube there are live captions in
33:03
YouTube there are live captions in English for the live stream events and
33:05
English for the live stream events and
33:05
English for the live stream events and then for all the on-demand talks that
33:08
then for all the on-demand talks that
33:08
then for all the on-demand talks that are already published as well as you
33:10
are already published as well as you
33:10
are already published as well as you know once these live talks are published
33:12
know once these live talks are published
33:12
know once these live talks are published within the next couple of weeks they're
33:14
within the next couple of weeks they're
33:14
within the next couple of weeks they're not only available but with English
33:15
not only available but with English
33:15
not only available but with English captions but we're going to have them
33:17
captions but we're going to have them
33:17
captions but we're going to have them those captions translated into a whole
33:19
those captions translated into a whole
33:19
those captions translated into a whole bunch of different languages
33:22
there you go
33:24
there you go
33:24
there you go um okay Discord the virtual hallway
33:27
um okay Discord the virtual hallway
33:27
um okay Discord the virtual hallway track
33:28
track
33:28
track what do we want to say Rob well I would
33:31
what do we want to say Rob well I would
33:31
what do we want to say Rob well I would say uh I I don't know how many folks
33:32
say uh I I don't know how many folks
33:32
say uh I I don't know how many folks joined us for the path to cytuscon the
33:34
joined us for the path to cytuscon the
33:34
joined us for the path to cytuscon the ones that did uh those events we had a
33:36
ones that did uh those events we had a
33:36
ones that did uh those events we had a great time talking on Discord uh and
33:38
great time talking on Discord uh and
33:39
great time talking on Discord uh and generally you know in a normal
33:40
generally you know in a normal
33:40
generally you know in a normal conference you know in person conference
33:43
conference you know in person conference
33:43
conference you know in person conference there's always that hallway track
33:44
there's always that hallway track
33:44
there's always that hallway track there's always some back Channel going
33:46
there's always some back Channel going
33:46
there's always some back Channel going on where you get to talk and sort of you
33:48
on where you get to talk and sort of you
33:48
on where you get to talk and sort of you know talk about what is in the talks but
33:50
know talk about what is in the talks but
33:50
know talk about what is in the talks but also what is going on in general uh and
33:53
also what is going on in general uh and
33:53
also what is going on in general uh and I think we're gonna have a good time in
33:54
I think we're gonna have a good time in
33:54
I think we're gonna have a good time in there today I will certainly be watching
33:56
there today I will certainly be watching
33:56
there today I will certainly be watching and I hope to talk to a bunch of you all
33:57
and I hope to talk to a bunch of you all
33:57
and I hope to talk to a bunch of you all there today yeah I'm going to be focused
34:00
there today yeah I'm going to be focused
34:00
there today yeah I'm going to be focused on the live stream during the live
34:01
on the live stream during the live
34:01
on the live stream during the live stream as co-host but as soon as this is
34:04
stream as co-host but as soon as this is
34:04
stream as co-host but as soon as this is over I'll be popping into the Discord
34:05
over I'll be popping into the Discord
34:06
over I'll be popping into the Discord and I I can't wait to you know be part
34:08
and I I can't wait to you know be part
34:08
and I I can't wait to you know be part of be part of that conversation um so
34:10
of be part of that conversation um so
34:10
of be part of that conversation um so it's the aka.ms slash open source
34:13
it's the aka.ms slash open source
34:13
it's the aka.ms slash open source Discord and it's the hashtag societoscon
34:16
Discord and it's the hashtag societoscon
34:16
Discord and it's the hashtag societoscon Channel once you're in there so please
34:17
Channel once you're in there so please
34:18
Channel once you're in there so please please join yeah and also if you're
34:20
please join yeah and also if you're
34:20
please join yeah and also if you're watching depending on where you're
34:21
watching depending on where you're
34:21
watching depending on where you're watching and how you're watching that's
34:23
watching and how you're watching that's
34:23
watching and how you're watching that's probably the best way to ask questions
34:24
probably the best way to ask questions
34:24
probably the best way to ask questions to the speakers so we'll also be
34:27
to the speakers so we'll also be
34:27
to the speakers so we'll also be watching in there to see what we can
34:28
watching in there to see what we can
34:28
watching in there to see what we can funnel to the speakers who are in the
34:29
funnel to the speakers who are in the
34:29
funnel to the speakers who are in the live stream today so definitely check it
34:31
live stream today so definitely check it
34:31
live stream today so definitely check it out
34:32
out
34:32
out um so here we're going to have six talks
34:34
um so here we're going to have six talks
34:34
um so here we're going to have six talks in today's live stream starting with our
34:36
in today's live stream starting with our
34:36
in today's live stream starting with our keynote speaker
34:37
keynote speaker
34:37
keynote speaker um Simon Willison and then there's just
34:39
um Simon Willison and then there's just
34:39
um Simon Willison and then there's just a fabulous collection of people with
34:42
a fabulous collection of people with
34:42
a fabulous collection of people with different postgres expertise so that's
34:44
different postgres expertise so that's
34:44
different postgres expertise so that's what we're here for but I just wanted to
34:46
what we're here for but I just wanted to
34:46
what we're here for but I just wanted to make sure jumping to the emea live
34:48
make sure jumping to the emea live
34:48
make sure jumping to the emea live stream speakers
34:49
stream speakers
34:49
stream speakers um that you all know that there is
34:52
um that you all know that there is
34:52
um that you all know that there is another completely different set of
34:54
another completely different set of
34:54
another completely different set of talks coming
34:55
talks coming
34:55
talks coming um and the URL that's showing in the
34:58
um and the URL that's showing in the
34:58
um and the URL that's showing in the upper right hand side of the screen um
35:00
upper right hand side of the screen um
35:00
upper right hand side of the screen um AKA dot Ms slash slidescon anea that's a
35:04
AKA dot Ms slash slidescon anea that's a
35:04
AKA dot Ms slash slidescon anea that's a calendar invite so if you want to drop
35:06
calendar invite so if you want to drop
35:06
calendar invite so if you want to drop it on your calendar block the time make
35:08
it on your calendar block the time make
35:08
it on your calendar block the time make sure that you're not double booked or
35:09
sure that you're not double booked or
35:09
sure that you're not double booked or something that's an easy way to do it
35:11
something that's an easy way to do it
35:11
something that's an easy way to do it and these folks are great but I know we
35:13
and these folks are great but I know we
35:13
and these folks are great but I know we need to get to Simon's talk so we
35:15
need to get to Simon's talk so we
35:15
need to get to Simon's talk so we probably can't talk about each of these
35:16
probably can't talk about each of these
35:16
probably can't talk about each of these amazing things right now
35:18
amazing things right now
35:18
amazing things right now um oh and there's more
35:22
oh yeah we've got a ton of talks today
35:24
oh yeah we've got a ton of talks today
35:24
oh yeah we've got a ton of talks today uh and and like we said definitely worth
35:27
uh and and like we said definitely worth
35:27
uh and and like we said definitely worth checking out check out the guide
35:29
checking out check out the guide
35:29
checking out check out the guide um and and we'll come back we'll talk
35:31
um and and we'll come back we'll talk
35:31
um and and we'll come back we'll talk about more about this later as the day
35:32
about more about this later as the day
35:32
about more about this later as the day goes on
35:34
goes on
35:34
goes on all right if you are posting on social I
35:36
all right if you are posting on social I
35:36
all right if you are posting on social I love it when people take photos or live
35:39
love it when people take photos or live
35:39
love it when people take photos or live tweet quotes or inspirational things
35:41
tweet quotes or inspirational things
35:41
tweet quotes or inspirational things that they learn from people's
35:43
that they learn from people's
35:43
that they learn from people's um presentations use the hashtag
35:45
um presentations use the hashtag
35:45
um presentations use the hashtag statuscon hashtag whatever social
35:46
statuscon hashtag whatever social
35:46
statuscon hashtag whatever social platform you're using and then for those
35:49
platform you're using and then for those
35:49
platform you're using and then for those of you who care about Swag
35:51
of you who care about Swag
35:51
of you who care about Swag um there is an opportunity to win these
35:53
um there is an opportunity to win these
35:53
um there is an opportunity to win these really cool Swag Bags um there's 75 of
35:56
really cool Swag Bags um there's 75 of
35:56
really cool Swag Bags um there's 75 of them being given away per live stream
35:57
them being given away per live stream
35:57
them being given away per live stream use the URL that's showing on the screen
36:00
use the URL that's showing on the screen
36:00
use the URL that's showing on the screen right now AKA dot Ms slash slightest con
36:03
right now AKA dot Ms slash slightest con
36:03
right now AKA dot Ms slash slightest con swag
36:04
swag
36:04
swag and the codes will get shared in the
36:07
and the codes will get shared in the
36:07
and the codes will get shared in the banners during each of the talks so
36:09
banners during each of the talks so
36:09
banners during each of the talks so you'll need a code to enter in
36:12
you'll need a code to enter in
36:12
you'll need a code to enter in um and the socks look pretty darn cool
36:15
um and the socks look pretty darn cool
36:15
um and the socks look pretty darn cool there's also sticker packs so and you
36:18
there's also sticker packs so and you
36:18
there's also sticker packs so and you can enter for both the swag bag and the
36:21
can enter for both the swag bag and the
36:21
can enter for both the swag bag and the sticker packs and there's 200 sticker
36:23
sticker packs and there's 200 sticker
36:23
sticker packs and there's 200 sticker packs with this collection being given
36:24
packs with this collection being given
36:24
packs with this collection being given away per live stream I don't know about
36:26
away per live stream I don't know about
36:26
away per live stream I don't know about you but I love stickers and I've got
36:28
you but I love stickers and I've got
36:28
you but I love stickers and I've got them all over my laptop Rob
36:30
them all over my laptop Rob
36:30
them all over my laptop Rob you know I have to admit I'm one of
36:32
you know I have to admit I'm one of
36:32
you know I have to admit I'm one of those clean laptop people so I don't
36:34
those clean laptop people so I don't
36:34
those clean laptop people so I don't typically put stickers on it but I'm
36:35
typically put stickers on it but I'm
36:35
typically put stickers on it but I'm starting a new thing which is putting
36:37
starting a new thing which is putting
36:37
starting a new thing which is putting stickers on my luggage uh so that it's
36:40
stickers on my luggage uh so that it's
36:40
stickers on my luggage uh so that it's more easy to spot that when I travel you
36:42
more easy to spot that when I travel you
36:42
more easy to spot that when I travel you know there's not as many uh postgres
36:44
know there's not as many uh postgres
36:44
know there's not as many uh postgres things out in the world on people's
36:46
things out in the world on people's
36:46
things out in the world on people's luggage but I think that's an easy way
36:47
luggage but I think that's an easy way
36:47
luggage but I think that's an easy way also to find people when you're
36:48
also to find people when you're
36:48
also to find people when you're traveling who might be into the things
36:50
traveling who might be into the things
36:50
traveling who might be into the things you're into so pretty good idea very
36:53
you're into so pretty good idea very
36:53
you're into so pretty good idea very retro Vibe too I like that okay I'm
36:56
retro Vibe too I like that okay I'm
36:56
retro Vibe too I like that okay I'm gonna I need to get myself some of these
36:58
gonna I need to get myself some of these
36:58
gonna I need to get myself some of these stickers too okay
36:59
stickers too okay
36:59
stickers too okay um there is going to be an attendee
37:01
um there is going to be an attendee
37:01
um there is going to be an attendee survey we this is the second annual
37:03
survey we this is the second annual
37:03
survey we this is the second annual cytuscon and event for postgres
37:05
cytuscon and event for postgres
37:05
cytuscon and event for postgres um we're gonna want your feedback if
37:07
um we're gonna want your feedback if
37:07
um we're gonna want your feedback if you're willing to give it to us so that
37:09
you're willing to give it to us so that
37:09
you're willing to give it to us so that we can make the event even better in the
37:11
we can make the event even better in the
37:11
we can make the event even better in the future
37:12
future
37:12
future um so there's a QR code there's a short
37:13
um so there's a QR code there's a short
37:13
um so there's a QR code there's a short URL on the screen
37:16
URL on the screen
37:16
URL on the screen um please and you can also give feedback
37:18
um please and you can also give feedback
37:18
um please and you can also give feedback on each and every talk and if you watch
37:20
on each and every talk and if you watch
37:20
on each and every talk and if you watch some talks today and some talks tomorrow
37:21
some talks today and some talks tomorrow
37:21
some talks today and some talks tomorrow you can come back and fill out the
37:23
you can come back and fill out the
37:23
you can come back and fill out the survey again and again before it closes
37:26
survey again and again before it closes
37:26
survey again and again before it closes on next Friday April 28th end of day
37:29
on next Friday April 28th end of day
37:29
on next Friday April 28th end of day anywhere are on earth time zone
37:32
anywhere are on earth time zone
37:32
anywhere are on earth time zone so that's important and finally did we
37:35
so that's important and finally did we
37:35
so that's important and finally did we talk about Discord already I don't think
37:37
talk about Discord already I don't think
37:37
talk about Discord already I don't think we did but uh we could talk about it
37:40
we did but uh we could talk about it
37:40
we did but uh we could talk about it more in the Discord if we were on the
37:41
more in the Discord if we were on the
37:41
more in the Discord if we were on the Discord and then we could talk about it
37:42
Discord and then we could talk about it
37:42
Discord and then we could talk about it with everybody who has joined the
37:44
with everybody who has joined the
37:44
with everybody who has joined the Discord definitely join the back channel
37:46
Discord definitely join the back channel
37:46
Discord definitely join the back channel it's it's it's a nice way to not just be
37:49
it's it's it's a nice way to not just be
37:49
it's it's it's a nice way to not just be watching the live stream but
37:50
watching the live stream but
37:50
watching the live stream but participating in the conversation too
37:53
participating in the conversation too
37:53
participating in the conversation too um so without further Ado I think that
37:56
um so without further Ado I think that
37:56
um so without further Ado I think that we should introduce our keynote speaker
37:58
we should introduce our keynote speaker
37:58
we should introduce our keynote speaker Simon Willison
38:01
Simon Willison
38:01
Simon Willison absolutely welcome Simon hi it's great
38:04
absolutely welcome Simon hi it's great
38:04
absolutely welcome Simon hi it's great to be here
38:05
to be here
38:05
to be here I'm so glad you're here
38:08
I'm so glad you're here
38:08
I'm so glad you're here um okay so for those of you who don't
38:10
um okay so for those of you who don't
38:10
um okay so for those of you who don't know Simon he's the co-creator of Django
38:12
know Simon he's the co-creator of Django
38:12
know Simon he's the co-creator of Django he's the creator of data set which
38:15
he's the creator of data set which
38:15
he's the creator of data set which you'll hear a bit more about today I've
38:18
you'll hear a bit more about today I've
38:18
you'll hear a bit more about today I've been following Simon for years on
38:19
been following Simon for years on
38:19
been following Simon for years on Twitter
38:20
Twitter
38:20
Twitter um Simon does a lot of his work in
38:22
um Simon does a lot of his work in
38:22
um Simon does a lot of his work in public which means that people like me
38:23
public which means that people like me
38:24
public which means that people like me can learn from his successes his
38:26
can learn from his successes his
38:26
can learn from his successes his failures his learnings
38:29
failures his learnings
38:29
failures his learnings um and it's I I've learned so much from
38:31
um and it's I I've learned so much from
38:31
um and it's I I've learned so much from you Simon these last couple of years
38:34
you Simon these last couple of years
38:34
you Simon these last couple of years I got to tell you I lost uh several
38:36
I got to tell you I lost uh several
38:36
I got to tell you I lost uh several hours the other day after the first half
38:38
hours the other day after the first half
38:38
hours the other day after the first half the sign has gone uh no doubt reading
38:40
the sign has gone uh no doubt reading
38:40
the sign has gone uh no doubt reading his things I learned blog which is kind
38:42
his things I learned blog which is kind
38:42
his things I learned blog which is kind of like a micro blog of sorts uh it was
38:44
of like a micro blog of sorts uh it was
38:44
of like a micro blog of sorts uh it was fantastic uh and you know just sort of
38:46
fantastic uh and you know just sort of
38:46
fantastic uh and you know just sort of reminded me uh I should really get back
38:48
reminded me uh I should really get back
38:48
reminded me uh I should really get back to doing more blogging it's really good
38:50
to doing more blogging it's really good
38:50
to doing more blogging it's really good stuff
38:52
stuff
38:52
stuff um yes I should point out that Simon was
38:54
um yes I should point out that Simon was
38:54
um yes I should point out that Simon was a guest on the first episode of path to
38:57
a guest on the first episode of path to
38:57
a guest on the first episode of path to side is gone which is this live audio
38:59
side is gone which is this live audio
38:59
side is gone which is this live audio show on Discord kind of like a podcast
39:00
show on Discord kind of like a podcast
39:00
show on Discord kind of like a podcast about a text chat and uh yeah we talked
39:03
about a text chat and uh yeah we talked
39:03
about a text chat and uh yeah we talked about working in public which is a whole
39:05
about working in public which is a whole
39:05
about working in public which is a whole fascinating conversation in and of
39:07
fascinating conversation in and of
39:07
fascinating conversation in and of itself but you're here to talk about big
39:09
itself but you're here to talk about big
39:09
itself but you're here to talk about big opportunities in small data
39:12
opportunities in small data
39:12
opportunities in small data so without further Ado I think we should
39:14
so without further Ado I think we should
39:14
so without further Ado I think we should pass the the floor to you you've got the
39:16
pass the the floor to you you've got the
39:16
pass the the floor to you you've got the stage signing
39:18
stage signing
39:18
stage signing fantastic so yeah today I want to talk
39:20
fantastic so yeah today I want to talk
39:20
fantastic so yeah today I want to talk about small data and um the reason I'm
39:23
about small data and um the reason I'm
39:23
about small data and um the reason I'm talking about that here is I feel like
39:25
talking about that here is I feel like
39:25
talking about that here is I feel like as an industry We've Got Big Data pretty
39:27
as an industry We've Got Big Data pretty
39:27
as an industry We've Got Big Data pretty much figured out at this point you know
39:29
much figured out at this point you know
39:29
much figured out at this point you know you've got tools like situs that let you
39:31
you've got tools like situs that let you
39:31
you've got tools like situs that let you horizontally scale your postcodes
39:32
horizontally scale your postcodes
39:32
horizontally scale your postcodes database it feels like every data
39:34
database it feels like every data
39:34
database it feels like every data warehouse product out there that I care
39:36
warehouse product out there that I care
39:36
warehouse product out there that I care about is is adding postgres um like
39:38
about is is adding postgres um like
39:38
about is is adding postgres um like protocol
39:40
protocol
39:40
protocol um protocol support and so forth like we
39:42
um protocol support and so forth like we
39:42
um protocol support and so forth like we know what to do if somebody has Peter
39:43
know what to do if somebody has Peter
39:43
know what to do if somebody has Peter bytes of data now that's not a difficult
39:45
bytes of data now that's not a difficult
39:45
bytes of data now that's not a difficult problem for us to solve but I'm really
39:47
problem for us to solve but I'm really
39:47
problem for us to solve but I'm really interested in the very other end of the
39:49
interested in the very other end of the
39:49
interested in the very other end of the scale the last um five years I've been
39:51
scale the last um five years I've been
39:51
scale the last um five years I've been exploring the issue of small data where
39:53
exploring the issue of small data where
39:53
exploring the issue of small data where I Define small as too big for Excel but
39:56
I Define small as too big for Excel but
39:56
I Define small as too big for Excel but small enough that it fits on a thumb
39:58
small enough that it fits on a thumb
39:58
small enough that it fits on a thumb drive or fits on your telephone my
40:00
drive or fits on your telephone my
40:00
drive or fits on your telephone my telephone's got a half terabyte of space
40:02
telephone's got a half terabyte of space
40:02
telephone's got a half terabyte of space on it right now so so small data can get
40:04
on it right now so so small data can get
40:04
on it right now so so small data can get pretty big
40:05
pretty big
40:05
pretty big um and the uh the way I've been
40:07
um and the uh the way I've been
40:07
um and the uh the way I've been exploring that is through an open source
40:08
exploring that is through an open source
40:08
exploring that is through an open source project that I've been working on called
40:10
project that I've been working on called
40:10
project that I've been working on called dataset
40:11
dataset
40:11
dataset um and I'm actually going rather than
40:13
um and I'm actually going rather than
40:13
um and I'm actually going rather than talk about what data set is I'm going to
40:14
talk about what data set is I'm going to
40:15
talk about what data set is I'm going to dive straight into a demo and show you
40:16
dive straight into a demo and show you
40:16
dive straight into a demo and show you what it is and what what it can do and
40:18
what it is and what what it can do and
40:18
what it is and what what it can do and how it works um this right here is the
40:22
how it works um this right here is the
40:22
how it works um this right here is the city of San Francisco's open data portal
40:24
city of San Francisco's open data portal
40:24
city of San Francisco's open data portal this is a trend from the past decade
40:26
this is a trend from the past decade
40:27
this is a trend from the past decade which I'm very excited about where local
40:29
which I'm very excited about where local
40:29
which I'm very excited about where local and National governments all around the
40:31
and National governments all around the
40:31
and National governments all around the world have been launching these these
40:33
world have been launching these these
40:33
world have been launching these these data portals where they publish data
40:35
data portals where they publish data
40:35
data portals where they publish data about the places where people live
40:37
about the places where people live
40:37
about the places where people live um and some of these things are updated
40:39
um and some of these things are updated
40:39
um and some of these things are updated on a daily basis like it's really
40:40
on a daily basis like it's really
40:40
on a daily basis like it's really interesting seeing quite how much
40:42
interesting seeing quite how much
40:42
interesting seeing quite how much information is flowing through all of
40:44
information is flowing through all of
40:44
information is flowing through all of these things the problem of course is
40:45
these things the problem of course is
40:45
these things the problem of course is that just because you published data
40:47
that just because you published data
40:47
that just because you published data doesn't mean that people can actually
40:48
doesn't mean that people can actually
40:48
doesn't mean that people can actually use it
40:49
use it
40:49
use it um and so in this case I'm going to do a
40:51
um and so in this case I'm going to do a
40:51
um and so in this case I'm going to do a demo using the city of San Francisco's
40:53
demo using the city of San Francisco's
40:53
demo using the city of San Francisco's City facilities data and this is
40:55
City facilities data and this is
40:55
City facilities data and this is definitely small data it's 1 700 rows
40:59
definitely small data it's 1 700 rows
40:59
definitely small data it's 1 700 rows right it's absolutely tiny and what I
41:01
right it's absolutely tiny and what I
41:01
right it's absolutely tiny and what I can do is I can copy and paste that URL
41:02
can do is I can copy and paste that URL
41:02
can do is I can copy and paste that URL and I'm going to load it into data set
41:04
and I'm going to load it into data set
41:04
and I'm going to load it into data set dataset is a python web application it
41:07
dataset is a python web application it
41:07
dataset is a python web application it um supports plugins which are similar to
41:09
um supports plugins which are similar to
41:09
um supports plugins which are similar to postgres extensions and I've got plugins
41:11
postgres extensions and I've got plugins
41:11
postgres extensions and I've got plugins installed here for things like load data
41:14
installed here for things like load data
41:14
installed here for things like load data from an open data portal so I'm going to
41:16
from an open data portal so I'm going to
41:16
from an open data portal so I'm going to fight this up here and paste in that URL
41:19
fight this up here and paste in that URL
41:19
fight this up here and paste in that URL I just pasted hit that that button and
41:21
I just pasted hit that that button and
41:21
I just pasted hit that that button and dataset will go and fetch that data and
41:24
dataset will go and fetch that data and
41:24
dataset will go and fetch that data and extract things like the metadata around
41:26
extract things like the metadata around
41:26
extract things like the metadata around it so what the columns are and it'll
41:28
it so what the columns are and it'll
41:28
it so what the columns are and it'll also pull in that table of 1700 rows now
41:31
also pull in that table of 1700 rows now
41:31
also pull in that table of 1700 rows now table's not particularly exciting but um
41:33
table's not particularly exciting but um
41:33
table's not particularly exciting but um one of the one of the data set plugins
41:35
one of the one of the data set plugins
41:35
one of the one of the data set plugins I'm running is it's called data set
41:37
I'm running is it's called data set
41:37
I'm running is it's called data set cluster map and it looks for data with a
41:39
cluster map and it looks for data with a
41:39
cluster map and it looks for data with a longitude and latitude column and sticks
41:42
longitude and latitude column and sticks
41:42
longitude and latitude column and sticks that data on a map so now I can click uh
41:46
that data on a map so now I can click uh
41:46
that data on a map so now I can click uh where are we I can click load all down
41:48
where are we I can click load all down
41:48
where are we I can click load all down at the bottom and now I've got all 1700
41:50
at the bottom and now I've got all 1700
41:50
at the bottom and now I've got all 1700 facilities loaded onto this map and
41:53
facilities loaded onto this map and
41:53
facilities loaded onto this map and already we're starting to see stories in
41:55
already we're starting to see stories in
41:55
already we're starting to see stories in the data like why does the city of San
41:58
the data like why does the city of San
41:58
the data like why does the city of San Francisco have 21 buildings down here it
42:00
Francisco have 21 buildings down here it
42:00
Francisco have 21 buildings down here it turns out this is a juvenile detention
42:01
turns out this is a juvenile detention
42:01
turns out this is a juvenile detention facility that the city owns and operates
42:04
facility that the city owns and operates
42:04
facility that the city owns and operates down in lahonda if you scroll to the
42:07
down in lahonda if you scroll to the
42:07
down in lahonda if you scroll to the side you'll see that there are places
42:08
side you'll see that there are places
42:08
side you'll see that there are places things like these this is um
42:10
things like these this is um
42:10
things like these this is um infrastructure for the hetchy reservoir
42:13
infrastructure for the hetchy reservoir
42:13
infrastructure for the hetchy reservoir this is a Heche substation and then over
42:15
this is a Heche substation and then over
42:15
this is a Heche substation and then over here we've got another cluster of
42:17
here we've got another cluster of
42:17
here we've got another cluster of markers
42:18
markers
42:18
markers um which is uh it's an it's a camp this
42:20
um which is uh it's an it's a camp this
42:20
um which is uh it's an it's a camp this is a a camp operated by The Parks and
42:22
is a a camp operated by The Parks and
42:22
is a a camp operated by The Parks and Recreation Department
42:24
Recreation Department
42:24
Recreation Department um an important feature of data set is
42:26
um an important feature of data set is
42:26
um an important feature of data set is what I call faceting it's essentially a
42:28
what I call faceting it's essentially a
42:28
what I call faceting it's essentially a group buying count so I can say you know
42:30
group buying count so I can say you know
42:30
group buying count so I can say you know what let's fast it by city and see that
42:32
what let's fast it by city and see that
42:32
what let's fast it by city and see that most of them are in San Francisco but
42:33
most of them are in San Francisco but
42:33
most of them are in San Francisco but you've got some in the Honda some in
42:35
you've got some in the Honda some in
42:35
you've got some in the Honda some in Groveland we're going to facet by
42:36
Groveland we're going to facet by
42:36
Groveland we're going to facet by jurisdiction now this is telling us
42:38
jurisdiction now this is telling us
42:38
jurisdiction now this is telling us which jurisdictions within the San
42:41
which jurisdictions within the San
42:41
which jurisdictions within the San Francisco government own the most
42:42
Francisco government own the most
42:42
Francisco government own the most property and Parks and Recreation the
42:44
property and Parks and Recreation the
42:44
property and Parks and Recreation the top Port is to 213 the airport's quite
42:48
top Port is to 213 the airport's quite
42:48
top Port is to 213 the airport's quite big the school district but some so just
42:51
big the school district but some so just
42:51
big the school district but some so just with a few clicks with importing that
42:52
with a few clicks with importing that
42:52
with a few clicks with importing that data into a new tool with start already
42:54
data into a new tool with start already
42:54
data into a new tool with start already starting to learn things about the data
42:57
starting to learn things about the data
42:57
starting to learn things about the data with a sort of aggregate level and this
42:59
with a sort of aggregate level and this
42:59
with a sort of aggregate level and this is something I'm really passionate about
43:00
is something I'm really passionate about
43:00
is something I'm really passionate about this idea that you can take data from a
43:03
this idea that you can take data from a
43:03
this idea that you can take data from a open data portal in basically any shape
43:05
open data portal in basically any shape
43:05
open data portal in basically any shape or size and quickly start diving into it
43:08
or size and quickly start diving into it
43:08
or size and quickly start diving into it and figuring okay what are the
43:09
and figuring okay what are the
43:09
and figuring okay what are the interesting Trends and patterns what
43:11
interesting Trends and patterns what
43:11
interesting Trends and patterns what does this look like if we visualize it
43:13
does this look like if we visualize it
43:13
does this look like if we visualize it um
43:13
um
43:13
um another feature of data set I should
43:15
another feature of data set I should
43:15
another feature of data set I should demonstrate quickly is all of this is
43:17
demonstrate quickly is all of this is
43:17
demonstrate quickly is all of this is running on top of a SQL database so when
43:20
running on top of a SQL database so when
43:20
running on top of a SQL database so when I filter by you know 38 for public
43:22
I filter by you know 38 for public
43:22
I filter by you know 38 for public health and then
43:24
health and then
43:24
health and then um 37 in San Francisco that's building
43:26
um 37 in San Francisco that's building
43:26
um 37 in San Francisco that's building up a SQL where query for me when I click
43:29
up a SQL where query for me when I click
43:29
up a SQL where query for me when I click View and edit SQL it actually shows me
43:32
View and edit SQL it actually shows me
43:32
View and edit SQL it actually shows me that SQL query and it lets me edit it so
43:35
that SQL query and it lets me edit it so
43:35
that SQL query and it lets me edit it so I can say you know what I don't need any
43:37
I can say you know what I don't need any
43:37
I can say you know what I don't need any of this stuff I'm going to keep the rest
43:39
of this stuff I'm going to keep the rest
43:39
of this stuff I'm going to keep the rest in there I can hit run SQL and now I'm
43:42
in there I can hit run SQL and now I'm
43:42
in there I can hit run SQL and now I'm getting back just that just the results
43:44
getting back just that just the results
43:44
getting back just that just the results of that particular query
43:46
of that particular query
43:46
of that particular query um and that's bookmarkable I can copy
43:48
um and that's bookmarkable I can copy
43:48
um and that's bookmarkable I can copy and paste this URL and send it to
43:49
and paste this URL and send it to
43:49
and paste this URL and send it to somebody else and they will see that
43:51
somebody else and they will see that
43:51
somebody else and they will see that same query that I'm seeing even more
43:53
same query that I'm seeing even more
43:53
same query that I'm seeing even more importantly you can get it back out as
43:54
importantly you can get it back out as
43:54
importantly you can get it back out as CSV or you can get it out as Json
43:57
CSV or you can get it out as Json
43:57
CSV or you can get it out as Json and this means that data set can serve
44:00
and this means that data set can serve
44:00
and this means that data set can serve as an integration layer for anything
44:02
as an integration layer for anything
44:02
as an integration layer for anything that can speak Json which is effectively
44:04
that can speak Json which is effectively
44:04
that can speak Json which is effectively everything these days so there's a lot
44:06
everything these days so there's a lot
44:06
everything these days so there's a lot of power in being able to run these
44:09
of power in being able to run these
44:09
of power in being able to run these arbitrary queries against a read-only
44:11
arbitrary queries against a read-only
44:11
arbitrary queries against a read-only database that's safe there's nothing no
44:13
database that's safe there's nothing no
44:13
database that's safe there's nothing no damage that you can do with this
44:14
damage that you can do with this
44:15
damage that you can do with this um but it gives you a very flexible and
44:16
um but it gives you a very flexible and
44:16
um but it gives you a very flexible and Powerful way of of remixing data and
44:19
Powerful way of of remixing data and
44:19
Powerful way of of remixing data and exporting it out
44:20
exporting it out
44:20
exporting it out um and and doing interesting things with
44:22
um and and doing interesting things with
44:22
um and and doing interesting things with that with it afterwards
44:26
um we'll switch back to the slides and
44:29
um we'll switch back to the slides and
44:29
um we'll switch back to the slides and where's my slide control gone sorry
44:32
where's my slide control gone sorry
44:32
where's my slide control gone sorry about this
44:34
about this
44:34
about this there we are
44:36
there we are
44:36
there we are um so the reason I got interested in
44:37
um so the reason I got interested in
44:37
um so the reason I got interested in this in the first place was
44:39
this in the first place was
44:39
this in the first place was um I started out actually in the realm
44:41
um I started out actually in the realm
44:41
um I started out actually in the realm of data journalism so the um I worked
44:45
of data journalism so the um I worked
44:45
of data journalism so the um I worked for the Guardian newspaper in London and
44:47
for the Guardian newspaper in London and
44:47
for the Guardian newspaper in London and we realized that our journalists were
44:48
we realized that our journalists were
44:48
we realized that our journalists were collecting all sorts of fascinating
44:50
collecting all sorts of fascinating
44:50
collecting all sorts of fascinating facts about the world in order to create
44:52
facts about the world in order to create
44:52
facts about the world in order to create infographics and maps for the newspaper
44:54
infographics and maps for the newspaper
44:54
infographics and maps for the newspaper and every time this happened we'd
44:55
and every time this happened we'd
44:55
and every time this happened we'd collect all of this data and then it
44:57
collect all of this data and then it
44:57
collect all of this data and then it would sit on a hard drive under
44:58
would sit on a hard drive under
44:58
would sit on a hard drive under somebody's desk and we decided that we'd
45:00
somebody's desk and we decided that we'd
45:00
somebody's desk and we decided that we'd start sharing that data with the wider
45:02
start sharing that data with the wider
45:02
start sharing that data with the wider world try and publish the data behind
45:04
world try and publish the data behind
45:04
world try and publish the data behind the stories anytime we put out one of
45:06
the stories anytime we put out one of
45:06
the stories anytime we put out one of these infographics so we started a thing
45:08
these infographics so we started a thing
45:08
these infographics so we started a thing called the data blog and it was a Blog
45:10
called the data blog and it was a Blog
45:10
called the data blog and it was a Blog of data that we had collected to to
45:12
of data that we had collected to to
45:12
of data that we had collected to to support the reporting that we were doing
45:13
support the reporting that we were doing
45:13
support the reporting that we were doing and at the time we ended up using Google
45:16
and at the time we ended up using Google
45:16
and at the time we ended up using Google Sheets as the mechanism for publishing
45:18
Sheets as the mechanism for publishing
45:18
Sheets as the mechanism for publishing the data because it was free and it
45:20
the data because it was free and it
45:20
the data because it was free and it already existed and we didn't have to
45:22
already existed and we didn't have to
45:22
already existed and we didn't have to build anything custom but it always
45:23
build anything custom but it always
45:24
build anything custom but it always frustrated me I always felt like there
45:25
frustrated me I always felt like there
45:25
frustrated me I always felt like there should be a better way of um of
45:28
should be a better way of um of
45:28
should be a better way of um of publishing data something that was open
45:29
publishing data something that was open
45:29
publishing data something that was open source and more powerful and more
45:31
source and more powerful and more
45:31
source and more powerful and more flexible and the question I asked myself
45:33
flexible and the question I asked myself
45:33
flexible and the question I asked myself was what's the best possible way of
45:35
was what's the best possible way of
45:35
was what's the best possible way of publishing this structure your data and
45:37
publishing this structure your data and
45:37
publishing this structure your data and that's where the idea for data set came
45:38
that's where the idea for data set came
45:39
that's where the idea for data set came from data set was basically my attempt
45:41
from data set was basically my attempt
45:41
from data set was basically my attempt at answering that question about the
45:42
at answering that question about the
45:42
at answering that question about the best possible way of publishing data
45:44
best possible way of publishing data
45:45
best possible way of publishing data that let people both explore it and also
45:47
that let people both explore it and also
45:47
that let people both explore it and also do Integrations and automate it and and
45:49
do Integrations and automate it and and
45:49
do Integrations and automate it and and and things like that
45:52
and things like that
45:52
and things like that um I hinted at this a moment ago this
45:53
um I hinted at this a moment ago this
45:53
um I hinted at this a moment ago this idea of being able to do read-only SQL
45:56
idea of being able to do read-only SQL
45:56
idea of being able to do read-only SQL queries via an API which I find amusing
45:59
queries via an API which I find amusing
45:59
queries via an API which I find amusing because for most web applications this
46:01
because for most web applications this
46:01
because for most web applications this would be considered a SQL injection
46:03
would be considered a SQL injection
46:03
would be considered a SQL injection attack this is a security hole that you
46:05
attack this is a security hole that you
46:05
attack this is a security hole that you must must be prevented at all costs with
46:07
must must be prevented at all costs with
46:07
must must be prevented at all costs with data set it's a documented feature and I
46:09
data set it's a documented feature and I
46:09
data set it's a documented feature and I get away with that because data set open
46:11
get away with that because data set open
46:11
get away with that because data set open it treats data as read-only so you can't
46:13
it treats data as read-only so you can't
46:13
it treats data as read-only so you can't damage it um it sets a time limit on the
46:16
damage it um it sets a time limit on the
46:16
damage it um it sets a time limit on the queries so you can't go too long with
46:17
queries so you can't go too long with
46:18
queries so you can't go too long with them
46:18
them
46:18
them um and it uh
46:22
and it gives you that that API access so
46:25
and it gives you that that API access so
46:25
and it gives you that that API access so that you can you can you can do
46:27
that you can you can you can do
46:27
that you can you can you can do interesting things on top of it I'll
46:29
interesting things on top of it I'll
46:29
interesting things on top of it I'll show you a um problem that I solved with
46:31
show you a um problem that I solved with
46:31
show you a um problem that I solved with this just the other day
46:33
this just the other day
46:33
this just the other day um this is my blog I've been blogging
46:35
um this is my blog I've been blogging
46:35
um this is my blog I've been blogging for nearly 21 years about old manner of
46:38
for nearly 21 years about old manner of
46:38
for nearly 21 years about old manner of different topics
46:39
different topics
46:39
different topics um I've got a tag Cloud which goes on
46:41
um I've got a tag Cloud which goes on
46:41
um I've got a tag Cloud which goes on forever with all of the different things
46:43
forever with all of the different things
46:43
forever with all of the different things that I've written about
46:44
that I've written about
46:44
that I've written about um but I've decided that it's 2023 The
46:46
um but I've decided that it's 2023 The
46:46
um but I've decided that it's 2023 The Cutting Edge of publication right now is
46:48
Cutting Edge of publication right now is
46:48
Cutting Edge of publication right now is an email newsletter like email
46:49
an email newsletter like email
46:49
an email newsletter like email newsletters are very much back again and
46:52
newsletters are very much back again and
46:52
newsletters are very much back again and I wanted to start doing that but I
46:53
I wanted to start doing that but I
46:53
I wanted to start doing that but I didn't want to have to do any additional
46:54
didn't want to have to do any additional
46:54
didn't want to have to do any additional work for it so since I've got this data
46:58
work for it so since I've got this data
46:58
work for it so since I've got this data in a database and I've got a data set
47:00
in a database and I've got a data set
47:00
in a database and I've got a data set instance that gives me the ability to
47:02
instance that gives me the ability to
47:02
instance that gives me the ability to query my own blog I figured I'd try and
47:05
query my own blog I figured I'd try and
47:05
query my own blog I figured I'd try and automate the process of constructing
47:07
automate the process of constructing
47:07
automate the process of constructing that newsletter so I built this utterly
47:09
that newsletter so I built this utterly
47:09
that newsletter so I built this utterly terrifying SQL query which pulls from
47:12
terrifying SQL query which pulls from
47:12
terrifying SQL query which pulls from all of my different content types and
47:13
all of my different content types and
47:13
all of my different content types and arranges them together into into a
47:15
arranges them together into into a
47:15
arranges them together into into a format like this and then I built this
47:18
format like this and then I built this
47:18
format like this and then I built this observable notebook um if you haven't
47:20
observable notebook um if you haven't
47:20
observable notebook um if you haven't played with observable yet it's a really
47:21
played with observable yet it's a really
47:21
played with observable yet it's a really interesting tool it's basically kind of
47:23
interesting tool it's basically kind of
47:23
interesting tool it's basically kind of like jupyter notebooks and python but in
47:25
like jupyter notebooks and python but in
47:25
like jupyter notebooks and python but in JavaScript and it makes it very easy to
47:28
JavaScript and it makes it very easy to
47:28
JavaScript and it makes it very easy to build custom applications that pull in
47:30
build custom applications that pull in
47:30
build custom applications that pull in data from different sources remix it and
47:33
data from different sources remix it and
47:33
data from different sources remix it and reformat it in different ways and so
47:35
reformat it in different ways and so
47:35
reformat it in different ways and so what I did is I um I took that
47:38
what I did is I um I took that
47:38
what I did is I um I took that terrifying SQL query I dropped it into
47:40
terrifying SQL query I dropped it into
47:40
terrifying SQL query I dropped it into the notebook right
47:42
the notebook right
47:42
the notebook right here where's my skipped implementation
47:45
here where's my skipped implementation
47:45
here where's my skipped implementation so I dropped that in I wrote a bunch of
47:47
so I dropped that in I wrote a bunch of
47:47
so I dropped that in I wrote a bunch of JavaScript to glue it together into HTML
47:49
JavaScript to glue it together into HTML
47:49
JavaScript to glue it together into HTML and all of that sort of stuff and then I
47:51
and all of that sort of stuff and then I
47:51
and all of that sort of stuff and then I had it output the HTML for my newsletter
47:54
had it output the HTML for my newsletter
47:54
had it output the HTML for my newsletter and actually that's I can click this to
47:56
and actually that's I can click this to
47:56
and actually that's I can click this to get rid of the things I've already sent
47:58
get rid of the things I've already sent
47:58
get rid of the things I've already sent out so right now if I was to send a
47:59
out so right now if I was to send a
47:59
out so right now if I was to send a newsletter it would have this content in
48:01
newsletter it would have this content in
48:01
newsletter it would have this content in and I added a button that copies and
48:03
and I added a button that copies and
48:03
and I added a button that copies and pastes the clipboard so now I can pop
48:06
pastes the clipboard so now I can pop
48:06
pastes the clipboard so now I can pop into sub stack and hit paste and my
48:09
into sub stack and hit paste and my
48:09
into sub stack and hit paste and my newsletter is is ready to send this is
48:12
newsletter is is ready to send this is
48:12
newsletter is is ready to send this is also kind of a hack against sub stack
48:14
also kind of a hack against sub stack
48:14
also kind of a hack against sub stack because substack don't offer an API
48:16
because substack don't offer an API
48:16
because substack don't offer an API there's no official automated way to
48:19
there's no official automated way to
48:19
there's no official automated way to create content for sub stack but it
48:20
create content for sub stack but it
48:20
create content for sub stack but it turns out copy and paste is kind of the
48:22
turns out copy and paste is kind of the
48:22
turns out copy and paste is kind of the universal integration method I've done
48:25
universal integration method I've done
48:25
universal integration method I've done quite a few things where something
48:26
quite a few things where something
48:26
quite a few things where something doesn't have an API but if you fiddle
48:29
doesn't have an API but if you fiddle
48:29
doesn't have an API but if you fiddle around with copy and paste just enough
48:30
around with copy and paste just enough
48:30
around with copy and paste just enough you can actually do a lot of automation
48:32
you can actually do a lot of automation
48:32
you can actually do a lot of automation on top of it so a sort of key hack here
48:34
on top of it so a sort of key hack here
48:34
on top of it so a sort of key hack here is to abuse sub Stack's ability to paste
48:37
is to abuse sub Stack's ability to paste
48:37
is to abuse sub Stack's ability to paste content to automate the creation of this
48:39
content to automate the creation of this
48:39
content to automate the creation of this and this works now I've got Simon
48:41
and this works now I've got Simon
48:41
and this works now I've got Simon willison's sub snack newsletter you can
48:43
willison's sub snack newsletter you can
48:43
willison's sub snack newsletter you can subscribe to it I send it out once or
48:44
subscribe to it I send it out once or
48:44
subscribe to it I send it out once or twice a week and it's been it's been
48:47
twice a week and it's been it's been
48:47
twice a week and it's been it's been working out pretty well I've been
48:48
working out pretty well I've been
48:48
working out pretty well I've been running it for a month and it's picked
48:49
running it for a month and it's picked
48:49
running it for a month and it's picked up a bunch of subscribers and it's been
48:51
up a bunch of subscribers and it's been
48:51
up a bunch of subscribers and it's been almost no work I probably spent a lot
48:54
almost no work I probably spent a lot
48:54
almost no work I probably spent a lot more time fiddling with my notebook than
48:55
more time fiddling with my notebook than
48:55
more time fiddling with my notebook than I did actually like writing any custom
48:57
I did actually like writing any custom
48:57
I did actually like writing any custom content for that newsletter
49:00
content for that newsletter
49:00
content for that newsletter and that's really the um that really
49:02
and that's really the um that really
49:02
and that's really the um that really goes to illustrate how useful it is to
49:05
goes to illustrate how useful it is to
49:05
goes to illustrate how useful it is to have data with in with an API that
49:07
have data with in with an API that
49:07
have data with in with an API that speaks SQL right SQL is a very powerful
49:10
speaks SQL right SQL is a very powerful
49:10
speaks SQL right SQL is a very powerful sort of domain-specific language for
49:12
sort of domain-specific language for
49:12
sort of domain-specific language for remixing and combining data and then
49:15
remixing and combining data and then
49:15
remixing and combining data and then well some people use things like graphql
49:17
well some people use things like graphql
49:17
well some people use things like graphql to give themselves a an API for for
49:20
to give themselves a an API for for
49:20
to give themselves a an API for for um for manipulating data in custom ways
49:22
um for manipulating data in custom ways
49:22
um for manipulating data in custom ways sql's graphql from the 70s right it's
49:25
sql's graphql from the 70s right it's
49:25
sql's graphql from the 70s right it's it's worked for a very long time it's a
49:27
it's worked for a very long time it's a
49:27
it's worked for a very long time it's a very robust way of of doing this sort of
49:30
very robust way of of doing this sort of
49:30
very robust way of of doing this sort of um semi-safe custom Integrations and I
49:32
um semi-safe custom Integrations and I
49:32
um semi-safe custom Integrations and I think there's a huge amount of power
49:33
think there's a huge amount of power
49:33
think there's a huge amount of power just in using SQL like that making
49:35
just in using SQL like that making
49:35
just in using SQL like that making things bookmarkable giving things Json
49:37
things bookmarkable giving things Json
49:37
things bookmarkable giving things Json endpoints that accept SQL there's a lot
49:39
endpoints that accept SQL there's a lot
49:39
endpoints that accept SQL there's a lot of really cool stuff that you can do
49:41
of really cool stuff that you can do
49:41
of really cool stuff that you can do with that
49:43
with that
49:43
with that okay
49:46
um but crucially data set is not built
49:49
um but crucially data set is not built
49:49
um but crucially data set is not built on top of postgres the um the data the
49:51
on top of postgres the um the data the
49:51
on top of postgres the um the data the database that I'm using under the hood
49:53
database that I'm using under the hood
49:53
database that I'm using under the hood for all of this is SQL Lite it's my
49:55
for all of this is SQL Lite it's my
49:55
for all of this is SQL Lite it's my second favorite database after postgres
49:57
second favorite database after postgres
49:57
second favorite database after postgres and uh the re there are a few different
49:59
and uh the re there are a few different
49:59
and uh the re there are a few different reasons that I'm using sqlite for this
50:02
reasons that I'm using sqlite for this
50:02
reasons that I'm using sqlite for this um sqlite claims to be one of the most
50:04
um sqlite claims to be one of the most
50:04
um sqlite claims to be one of the most widely installed pieces of software in
50:06
widely installed pieces of software in
50:06
widely installed pieces of software in the world and absolutely the most widely
50:08
the world and absolutely the most widely
50:08
the world and absolutely the most widely deployed database engine it's in every
50:10
deployed database engine it's in every
50:10
deployed database engine it's in every iPhone every Android phone I'm pretty
50:12
iPhone every Android phone I'm pretty
50:12
iPhone every Android phone I'm pretty sure it's running on my Apple watch and
50:14
sure it's running on my Apple watch and
50:14
sure it's running on my Apple watch and counting my steps because sqlite is
50:16
counting my steps because sqlite is
50:16
counting my steps because sqlite is designed as an embedded Library it's a
50:19
designed as an embedded Library it's a
50:19
designed as an embedded Library it's a very very small very tightly written C
50:21
very very small very tightly written C
50:21
very very small very tightly written C library that you can drop into literally
50:24
library that you can drop into literally
50:24
library that you can drop into literally anything on any platform and it gives
50:26
anything on any platform and it gives
50:26
anything on any platform and it gives you the part the full power of a
50:27
you the part the full power of a
50:27
you the part the full power of a relational database
50:29
relational database
50:29
relational database um a crucial characteristic is that a
50:31
um a crucial characteristic is that a
50:31
um a crucial characteristic is that a sqlite database is just a file it's a
50:34
sqlite database is just a file it's a
50:34
sqlite database is just a file it's a data.db file that sits there on your
50:36
data.db file that sits there on your
50:36
data.db file that sits there on your disk and this makes them
50:38
disk and this makes them
50:38
disk and this makes them super easy to work with you can back
50:40
super easy to work with you can back
50:40
super easy to work with you can back them up by creating a copy you can share
50:42
them up by creating a copy you can share
50:42
them up by creating a copy you can share them by emailing them to people you're
50:44
them by emailing them to people you're
50:44
them by emailing them to people you're creating a new file is as expensive as
50:46
creating a new file is as expensive as
50:46
creating a new file is as expensive as creating any file on disk if you've lost
50:48
creating any file on disk if you've lost
50:48
creating any file on disk if you've lost interest in it you can just throw it
50:49
interest in it you can just throw it
50:49
interest in it you can just throw it away again so it's a very quick and
50:51
away again so it's a very quick and
50:51
away again so it's a very quick and agile way of working with data but
50:54
agile way of working with data but
50:54
agile way of working with data but despite the fact that it's so sort of
50:56
despite the fact that it's so sort of
50:56
despite the fact that it's so sort of like tiny and flexible it's also a very
50:58
like tiny and flexible it's also a very
50:58
like tiny and flexible it's also a very stable file format the um the Library of
51:01
stable file format the um the Library of
51:01
stable file format the um the Library of Congress lists sqlite as one of their
51:03
Congress lists sqlite as one of their
51:03
Congress lists sqlite as one of their approved archival mechanisms for
51:05
approved archival mechanisms for
51:05
approved archival mechanisms for structured data and that's because the
51:07
structured data and that's because the
51:07
structured data and that's because the sqlite team are meticulous about not
51:10
sqlite team are meticulous about not
51:10
sqlite team are meticulous about not breaking backwards compatibility when
51:12
breaking backwards compatibility when
51:12
breaking backwards compatibility when they when they release new versions so
51:13
they when they release new versions so
51:13
they when they release new versions so once your data is in sqlite it's going
51:15
once your data is in sqlite it's going
51:15
once your data is in sqlite it's going to basically be safe forever as
51:17
to basically be safe forever as
51:17
to basically be safe forever as something that you can then query and
51:19
something that you can then query and
51:19
something that you can then query and use in the future and that's that's as
51:21
use in the future and that's that's as
51:21
use in the future and that's that's as somebody who cares about newspapers
51:22
somebody who cares about newspapers
51:22
somebody who cares about newspapers that's a really big deal
51:24
that's a really big deal
51:24
that's a really big deal and then the other idea um that's the
51:27
and then the other idea um that's the
51:27
and then the other idea um that's the data set illustrates which um which came
51:29
data set illustrates which um which came
51:29
data set illustrates which um which came from from this from music which inspired
51:31
from from this from music which inspired
51:31
from from this from music which inspired me to use sqlite is something that I
51:33
me to use sqlite is something that I
51:33
me to use sqlite is something that I call the baked data architectural
51:35
call the baked data architectural
51:35
call the baked data architectural pattern these days deploying a web
51:38
pattern these days deploying a web
51:38
pattern these days deploying a web application has never been cheaper in
51:40
application has never been cheaper in
51:40
application has never been cheaper in terms of running that web application
51:42
terms of running that web application
51:42
terms of running that web application somewhere in the cloud you've got
51:44
somewhere in the cloud you've got
51:44
somewhere in the cloud you've got technology like kubernetes and Docker
51:45
technology like kubernetes and Docker
51:45
technology like kubernetes and Docker that there are hosting providers like
51:47
that there are hosting providers like
51:47
that there are hosting providers like AWS Lambda or Google Cloud run or the
51:50
AWS Lambda or Google Cloud run or the
51:50
AWS Lambda or Google Cloud run or the cell or fly who will all run a stateless
51:53
cell or fly who will all run a stateless
51:53
cell or fly who will all run a stateless web application for cents a month if
51:56
web application for cents a month if
51:56
web application for cents a month if it's not getting much traffic they're
51:58
it's not getting much traffic they're
51:58
it's not getting much traffic they're incredibly inexpensive there is a catch
52:00
incredibly inexpensive there is a catch
52:00
incredibly inexpensive there is a catch and the catch is that you can't run a
52:02
and the catch is that you can't run a
52:02
and the catch is that you can't run a database in them because all the reason
52:04
database in them because all the reason
52:04
database in them because all the reason these things are so cheap is that
52:06
these things are so cheap is that
52:06
these things are so cheap is that they're essentially read-only stateless
52:08
they're essentially read-only stateless
52:08
they're essentially read-only stateless web applications that you don't get to
52:10
web applications that you don't get to
52:10
web applications that you don't get to write to a disk or if you do write to a
52:12
write to a disk or if you do write to a
52:12
write to a disk or if you do write to a disk it doesn't get persisted anywhere
52:14
disk it doesn't get persisted anywhere
52:14
disk it doesn't get persisted anywhere which makes them very cheap and
52:15
which makes them very cheap and
52:15
which makes them very cheap and expensive to run but if you want to run
52:17
expensive to run but if you want to run
52:17
expensive to run but if you want to run a relational database you're kind of out
52:19
a relational database you're kind of out
52:19
a relational database you're kind of out of luck you have to pay a lot of extra
52:20
of luck you have to pay a lot of extra
52:20
of luck you have to pay a lot of extra money for a hosted database somewhere
52:22
money for a hosted database somewhere
52:22
money for a hosted database somewhere the thing I realized is that if your
52:25
the thing I realized is that if your
52:25
the thing I realized is that if your data is read only none of this matters
52:28
data is read only none of this matters
52:28
data is read only none of this matters like once you start dealing with
52:30
like once you start dealing with
52:30
like once you start dealing with read-only data and in my case it's
52:31
read-only data and in my case it's
52:31
read-only data and in my case it's sqlite database files so it's literally
52:33
sqlite database files so it's literally
52:33
sqlite database files so it's literally a blob of bytes on a disk if it's reader
52:36
a blob of bytes on a disk if it's reader
52:36
a blob of bytes on a disk if it's reader only the fact that you can't write back
52:38
only the fact that you can't write back
52:38
only the fact that you can't write back to the disk again isn't actually an
52:40
to the disk again isn't actually an
52:40
to the disk again isn't actually an issue and for the world that I care
52:41
issue and for the world that I care
52:41
issue and for the world that I care about coming from things like journalism
52:43
about coming from things like journalism
52:43
about coming from things like journalism a lot of the data sets that we're
52:45
a lot of the data sets that we're
52:45
a lot of the data sets that we're dealing with don't ever get updated you
52:47
dealing with don't ever get updated you
52:47
dealing with don't ever get updated you know it's a snapshot of San Francisco
52:48
know it's a snapshot of San Francisco
52:48
know it's a snapshot of San Francisco city facilities that might be updated a
52:51
city facilities that might be updated a
52:51
city facilities that might be updated a few times a month but it's not something
52:52
few times a month but it's not something
52:52
few times a month but it's not something that's going to accept constant rights
52:54
that's going to accept constant rights
52:54
that's going to accept constant rights so when you've got a read-only database
52:57
so when you've got a read-only database
52:57
so when you've got a read-only database there's a trick you can pull where you
52:59
there's a trick you can pull where you
52:59
there's a trick you can pull where you can effectively bundle that database up
53:01
can effectively bundle that database up
53:01
can effectively bundle that database up as part of the application that you're
53:03
as part of the application that you're
53:03
as part of the application that you're deploying onto these platforms and now
53:05
deploying onto these platforms and now
53:05
deploying onto these platforms and now you're now you can deploy a fully
53:08
you're now you can deploy a fully
53:08
you're now you can deploy a fully capable relational database interface
53:10
capable relational database interface
53:10
capable relational database interface that costs you a few cents a month in
53:13
that costs you a few cents a month in
53:13
that costs you a few cents a month in hosting costs um all of these pla or a
53:16
hosting costs um all of these pla or a
53:16
hosting costs um all of these pla or a lot of these platforms feature have a
53:17
lot of these platforms feature have a
53:17
lot of these platforms feature have a feature called scale to zero which means
53:20
feature called scale to zero which means
53:20
feature called scale to zero which means if there's no traffic coming into your
53:22
if there's no traffic coming into your
53:22
if there's no traffic coming into your application it doesn't even well it
53:23
application it doesn't even well it
53:24
application it doesn't even well it doesn't run it doesn't cost you any
53:25
doesn't run it doesn't cost you any
53:25
doesn't run it doesn't cost you any money at all the application is
53:26
money at all the application is
53:26
money at all the application is essentially static then they start the
53:28
essentially static then they start the
53:28
essentially static then they start the thing up running when the first HTTP
53:30
thing up running when the first HTTP
53:30
thing up running when the first HTTP request comes in when you've got scale
53:32
request comes in when you've got scale
53:32
request comes in when you've got scale to zero it means that your costs for
53:34
to zero it means that your costs for
53:34
to zero it means that your costs for these sort of low traffic projects uh
53:36
these sort of low traffic projects uh
53:36
these sort of low traffic projects uh effectively nothing I think um many of
53:39
effectively nothing I think um many of
53:39
effectively nothing I think um many of these providers here have a free tier
53:41
these providers here have a free tier
53:41
these providers here have a free tier and it's very common for projects that
53:43
and it's very common for projects that
53:43
and it's very common for projects that get little traffic to fit entirely
53:45
get little traffic to fit entirely
53:45
get little traffic to fit entirely within that tier so effectively we have
53:47
within that tier so effectively we have
53:47
within that tier so effectively we have free hosting for these applications if
53:49
free hosting for these applications if
53:49
free hosting for these applications if you can get away with just having
53:50
you can get away with just having
53:50
you can get away with just having read-only data and the flip side of that
53:53
read-only data and the flip side of that
53:53
read-only data and the flip side of that is that they scale up really well as
53:55
is that they scale up really well as
53:55
is that they scale up really well as well
53:56
well
53:56
well um if you want to serve millions of
53:57
um if you want to serve millions of
53:57
um if you want to serve millions of requests a second using a using this
53:59
requests a second using a using this
53:59
requests a second using a using this baked data pattern you can do that by
54:01
baked data pattern you can do that by
54:01
baked data pattern you can do that by running multiple copies spin up as many
54:04
running multiple copies spin up as many
54:04
running multiple copies spin up as many copies of the entire application and a
54:06
copies of the entire application and a
54:06
copies of the entire application and a full copy of the database in each one
54:07
full copy of the database in each one
54:07
full copy of the database in each one stick them behind a load balancer and
54:09
stick them behind a load balancer and
54:09
stick them behind a load balancer and you can handle as much traffic as you
54:11
you can handle as much traffic as you
54:11
you can handle as much traffic as you can throw at it which again for
54:12
can throw at it which again for
54:12
can throw at it which again for journalism when you have things like um
54:14
journalism when you have things like um
54:14
journalism when you have things like um like election results days and so forth
54:16
like election results days and so forth
54:16
like election results days and so forth this is a really useful useful ability
54:17
this is a really useful useful ability
54:17
this is a really useful useful ability as well
54:19
as well
54:19
as well I touched on this a little bit but I
54:21
I touched on this a little bit but I
54:21
I touched on this a little bit but I want to dive a bit more into the fact
54:22
want to dive a bit more into the fact
54:22
want to dive a bit more into the fact that SQL plus HTTP is a fantastic
54:26
that SQL plus HTTP is a fantastic
54:26
that SQL plus HTTP is a fantastic integration tool um and I'll show you
54:28
integration tool um and I'll show you
54:28
integration tool um and I'll show you another example of a project that I've
54:30
another example of a project that I've
54:30
another example of a project that I've run and this is a website called Niche
54:32
run and this is a website called Niche
54:32
run and this is a website called Niche hyphen museums.com and it's the website
54:35
hyphen museums.com and it's the website
54:35
hyphen museums.com and it's the website they ran for my main hobby which is
54:38
they ran for my main hobby which is
54:38
they ran for my main hobby which is seeking out and exploring tiny museums
54:41
seeking out and exploring tiny museums
54:41
seeking out and exploring tiny museums the idea behind this is um if you go to
54:43
the idea behind this is um if you go to
54:43
the idea behind this is um if you go to a big Museum sure it'll be interesting
54:45
a big Museum sure it'll be interesting
54:45
a big Museum sure it'll be interesting if you go to a really small Museum it
54:47
if you go to a really small Museum it
54:47
if you go to a really small Museum it doesn't matter what it's about the
54:49
doesn't matter what it's about the
54:49
doesn't matter what it's about the chances are that the person who is at
54:51
chances are that the person who is at
54:51
chances are that the person who is at the desk at that museum is the person
54:52
the desk at that museum is the person
54:52
the desk at that museum is the person who set it up so now you get to have a
54:54
who set it up so now you get to have a
54:54
who set it up so now you get to have a conversation with somebody who collects
54:56
conversation with somebody who collects
54:56
conversation with somebody who collects Pez dispensers or runs the Bigfoot
54:58
Pez dispensers or runs the Bigfoot
54:58
Pez dispensers or runs the Bigfoot Discovery Museum or in this case this is
55:00
Discovery Museum or in this case this is
55:00
Discovery Museum or in this case this is the misalignment museum in San Francisco
55:03
the misalignment museum in San Francisco
55:03
the misalignment museum in San Francisco which tells the story of a um it's it's
55:05
which tells the story of a um it's it's
55:05
which tells the story of a um it's it's an apology from AI for destroying
55:07
an apology from AI for destroying
55:07
an apology from AI for destroying Humanity in the future it's only open
55:10
Humanity in the future it's only open
55:10
Humanity in the future it's only open for a few more weeks it's worth worth
55:11
for a few more weeks it's worth worth
55:11
for a few more weeks it's worth worth popping in if you're in the area but
55:13
popping in if you're in the area but
55:13
popping in if you're in the area but anyway this website is actually just
55:15
anyway this website is actually just
55:15
anyway this website is actually just data set this is my data set web
55:17
data set this is my data set web
55:17
data set this is my data set web application with a custom index. HTML
55:20
application with a custom index. HTML
55:20
application with a custom index. HTML template so that it looks like a website
55:22
template so that it looks like a website
55:22
template so that it looks like a website and all of the features on this like use
55:23
and all of the features on this like use
55:24
and all of the features on this like use my location and search and the the the
55:26
my location and search and the the the
55:26
my location and search and the the the atom feed up here those are all just SQL
55:28
atom feed up here those are all just SQL
55:28
atom feed up here those are all just SQL queries that are baked into data set
55:30
queries that are baked into data set
55:30
queries that are baked into data set this right here is the query for the
55:33
this right here is the query for the
55:33
this right here is the query for the atom feed and um
55:34
atom feed and um
55:34
atom feed and um if we if we highlight these options here
55:37
if we if we highlight these options here
55:37
if we if we highlight these options here you can see that the way this works it
55:38
you can see that the way this works it
55:38
you can see that the way this works it selects things and it aliases them as
55:40
selects things and it aliases them as
55:40
selects things and it aliases them as atom ID atom title atom updated atom
55:43
atom ID atom title atom updated atom
55:43
atom ID atom title atom updated atom link then there's a data set plugin
55:45
link then there's a data set plugin
55:45
link then there's a data set plugin called data set atom which looks for
55:47
called data set atom which looks for
55:48
called data set atom which looks for those column names and if it finds them
55:49
those column names and if it finds them
55:49
those column names and if it finds them it produces a link to an atom feed so
55:52
it produces a link to an atom feed so
55:52
it produces a link to an atom feed so this single SQL query here defines the
55:55
this single SQL query here defines the
55:55
this single SQL query here defines the atom feed for the website purely by
55:57
atom feed for the website purely by
55:57
atom feed for the website purely by reshaping the data in SQL to that
56:00
reshaping the data in SQL to that
56:00
reshaping the data in SQL to that specific set of columns that terrifying
56:02
specific set of columns that terrifying
56:02
specific set of columns that terrifying coalesce chunk at the bottom is showing
56:04
coalesce chunk at the bottom is showing
56:04
coalesce chunk at the bottom is showing how D how in depth you can get with this
56:06
how D how in depth you can get with this
56:06
how D how in depth you can get with this if you want to construct in this case
56:08
if you want to construct in this case
56:08
if you want to construct in this case the HTML for the blog entry and this
56:11
the HTML for the blog entry and this
56:11
the HTML for the blog entry and this works right this is um net newswire the
56:13
works right this is um net newswire the
56:13
works right this is um net newswire the feedreader subscribed to the feed of
56:15
feedreader subscribed to the feed of
56:15
feedreader subscribed to the feed of content from my Niche Museum's website
56:17
content from my Niche Museum's website
56:17
content from my Niche Museum's website so there's a lot of different moving
56:19
so there's a lot of different moving
56:19
so there's a lot of different moving Parts involved here but fundamentally
56:21
Parts involved here but fundamentally
56:21
Parts involved here but fundamentally I've used a SQL query to transform data
56:23
I've used a SQL query to transform data
56:24
I've used a SQL query to transform data from whatever format it's stored in that
56:26
from whatever format it's stored in that
56:26
from whatever format it's stored in that in that website into a format that's
56:27
in that website into a format that's
56:27
in that website into a format that's compatible with feed readers and then
56:29
compatible with feed readers and then
56:29
compatible with feed readers and then they've got a little plugin that turns
56:30
they've got a little plugin that turns
56:30
they've got a little plugin that turns that into XML
56:33
that into XML
56:33
that into XML so the bake data pattern I've done a lot
56:35
so the bake data pattern I've done a lot
56:35
so the bake data pattern I've done a lot of exploration of this using sqlite a
56:38
of exploration of this using sqlite a
56:38
of exploration of this using sqlite a open question for me
56:40
open question for me
56:40
open question for me um is could you do this with postgres
56:42
um is could you do this with postgres
56:42
um is could you do this with postgres itself I'm very confident that this
56:45
itself I'm very confident that this
56:45
itself I'm very confident that this would work I think there's no reason at
56:47
would work I think there's no reason at
56:47
would work I think there's no reason at all you couldn't put together a Docker
56:49
all you couldn't put together a Docker
56:49
all you couldn't put together a Docker container that contains a full
56:50
container that contains a full
56:50
container that contains a full installation of post squares with data
56:52
installation of post squares with data
56:52
installation of post squares with data with data volumes and all of the stuff
56:54
with data volumes and all of the stuff
56:54
with data volumes and all of the stuff that postgres needs to serve a read-only
56:56
that postgres needs to serve a read-only
56:56
that postgres needs to serve a read-only database bundle that up with an
56:57
database bundle that up with an
56:57
database bundle that up with an application like a Django app or
56:59
application like a Django app or
56:59
application like a Django app or something on top and deploy the entire
57:01
something on top and deploy the entire
57:01
something on top and deploy the entire thing in exactly the same way as I'm
57:03
thing in exactly the same way as I'm
57:03
thing in exactly the same way as I'm doing with sqlite and this would give
57:05
doing with sqlite and this would give
57:05
doing with sqlite and this would give you all of those same benefits you could
57:07
you all of those same benefits you could
57:07
you all of those same benefits you could get a scale to zero thing where your
57:08
get a scale to zero thing where your
57:08
get a scale to zero thing where your website costs nothing to nothing at all
57:11
website costs nothing to nothing at all
57:11
website costs nothing to nothing at all if nobody's visiting it you could scale
57:13
if nobody's visiting it you could scale
57:13
if nobody's visiting it you could scale it to handle a million requests a second
57:14
it to handle a million requests a second
57:14
it to handle a million requests a second if um if you need to just by deploying
57:17
if um if you need to just by deploying
57:17
if um if you need to just by deploying multiple copies of that same thing so I
57:19
multiple copies of that same thing so I
57:19
multiple copies of that same thing so I think that's something very interesting
57:21
think that's something very interesting
57:21
think that's something very interesting and that could be explored around this
57:22
and that could be explored around this
57:22
and that could be explored around this and many other databases as well
57:24
and many other databases as well
57:24
and many other databases as well elasticsearch for example is something
57:26
elasticsearch for example is something
57:26
elasticsearch for example is something that you could absolutely bake into a
57:28
that you could absolutely bake into a
57:28
that you could absolutely bake into a container with its index data and get
57:30
container with its index data and get
57:30
container with its index data and get you know a very powerful search feature
57:32
you know a very powerful search feature
57:32
you know a very powerful search feature that costs to you nothing at all if
57:34
that costs to you nothing at all if
57:34
that costs to you nothing at all if nobody's using that particular
57:35
nobody's using that particular
57:35
nobody's using that particular application
57:37
application
57:37
application I've got one last demo
57:39
I've got one last demo
57:39
I've got one last demo um this again very much fits the theme
57:41
um this again very much fits the theme
57:41
um this again very much fits the theme of small data it turns out you can run
57:44
of small data it turns out you can run
57:44
of small data it turns out you can run things like data set entirely in your
57:46
things like data set entirely in your
57:46
things like data set entirely in your browser these days thanks to webassembly
57:48
browser these days thanks to webassembly
57:48
browser these days thanks to webassembly so I'm going to switch over to My Demo
57:50
so I'm going to switch over to My Demo
57:50
so I'm going to switch over to My Demo again
57:51
again
57:51
again and show you a thing that I built called
57:54
and show you a thing that I built called
57:54
and show you a thing that I built called data set light could we switch to the
57:57
data set light could we switch to the
57:57
data set light could we switch to the demo feed
57:59
demo feed
57:59
demo feed um oh my fault uh my my screen sharing
58:02
um oh my fault uh my my screen sharing
58:02
um oh my fault uh my my screen sharing stopped I'll just fix that now window
58:05
stopped I'll just fix that now window
58:05
stopped I'll just fix that now window uh
58:07
uh
58:07
uh here we go
58:09
here we go
58:09
here we go okay we should be there we go so um data
58:12
okay we should be there we go so um data
58:12
okay we should be there we go so um data set light is data set running entirely
58:15
set light is data set running entirely
58:15
set light is data set running entirely in webassembly so I can I've opened up
58:18
in webassembly so I can I've opened up
58:18
in webassembly so I can I've opened up the um the network pane here so you can
58:21
the um the network pane here so you can
58:21
the um the network pane here so you can see what it does in the left hand side
58:23
see what it does in the left hand side
58:23
see what it does in the left hand side it's loading up data set and installing
58:24
it's loading up data set and installing
58:24
it's loading up data set and installing packages and so forth on the right hand
58:26
packages and so forth on the right hand
58:26
packages and so forth on the right hand side you can see it downloaded something
58:27
side you can see it downloaded something
58:27
side you can see it downloaded something called
58:28
called
58:29
called piadite.asm.js that's a and piad idle
58:31
piadite.asm.js that's a and piad idle
58:31
piadite.asm.js that's a and piad idle asm.data this is a full python
58:34
asm.data this is a full python
58:34
asm.data this is a full python interpreter and the python standard
58:36
interpreter and the python standard
58:36
interpreter and the python standard Library bundled up for webassembly so
58:38
Library bundled up for webassembly so
58:38
Library bundled up for webassembly so that it runs in the browser so right
58:40
that it runs in the browser so right
58:40
that it runs in the browser so right here I have a server-side web
58:42
here I have a server-side web
58:42
here I have a server-side web application like data set is a
58:44
application like data set is a
58:44
application like data set is a traditional server side app that's
58:46
traditional server side app that's
58:46
traditional server side app that's running entirely client-side and most of
58:48
running entirely client-side and most of
58:48
running entirely client-side and most of the functionality is working I can facet
58:50
the functionality is working I can facet
58:50
the functionality is working I can facet and I can run SQL queries and all of
58:52
and I can run SQL queries and all of
58:52
and I can run SQL queries and all of those kinds of things
58:54
those kinds of things
58:54
those kinds of things um and I built this actually as a joke I
58:57
um and I built this actually as a joke I
58:57
um and I built this actually as a joke I thought it would be amusing to show that
58:59
thought it would be amusing to show that
58:59
thought it would be amusing to show that a server-side web app can run
59:00
a server-side web app can run
59:00
a server-side web app can run client-side now but I didn't think it
59:02
client-side now but I didn't think it
59:02
client-side now but I didn't think it would be useful for anything but since I
59:04
would be useful for anything but since I
59:04
would be useful for anything but since I built it I found that I'm using it for
59:06
built it I found that I'm using it for
59:06
built it I found that I'm using it for all sorts of different purposes mainly
59:08
all sorts of different purposes mainly
59:08
all sorts of different purposes mainly because it's 100 robust right there is
59:11
because it's 100 robust right there is
59:11
because it's 100 robust right there is no server this is all just HTML and
59:13
no server this is all just HTML and
59:13
no server this is all just HTML and JavaScript there is nothing that can
59:14
JavaScript there is nothing that can
59:14
JavaScript there is nothing that can break with this which means that I can
59:17
break with this which means that I can
59:17
break with this which means that I can build things on it and share them with
59:18
build things on it and share them with
59:18
build things on it and share them with people and not have to to worry about
59:19
people and not have to to worry about
59:19
people and not have to to worry about hosting Bills or the thing might broken
59:21
hosting Bills or the thing might broken
59:21
hosting Bills or the thing might broken 10 years time or anything like that so
59:24
10 years time or anything like that so
59:24
10 years time or anything like that so just yesterday a project called red
59:26
just yesterday a project called red
59:26
just yesterday a project called red pajama released a training data set for
59:29
pajama released a training data set for
59:30
pajama released a training data set for building chat GPT style language models
59:33
building chat GPT style language models
59:33
building chat GPT style language models um there is it was over what they it's
59:34
um there is it was over what they it's
59:34
um there is it was over what they it's 1.2 trillion tokens
59:37
1.2 trillion tokens
59:37
1.2 trillion tokens and this is how they released it it's a
59:39
and this is how they released it it's a
59:39
and this is how they released it it's a file called urls.txt with 2048 URLs in
59:44
file called urls.txt with 2048 URLs in
59:44
file called urls.txt with 2048 URLs in it each of these is like a gigabyte
59:47
it each of these is like a gigabyte
59:47
it each of these is like a gigabyte large file which you can then download
59:49
large file which you can then download
59:49
large file which you can then download and to gather all of this stuff to
59:50
and to gather all of this stuff to
59:50
and to gather all of this stuff to download all of it you'd need 2.7
59:52
download all of it you'd need 2.7
59:52
download all of it you'd need 2.7 terabytes of disk space I do not have
59:55
terabytes of disk space I do not have
59:55
terabytes of disk space I do not have 2.7 terabytes of disk space but I still
59:57
2.7 terabytes of disk space but I still
59:57
2.7 terabytes of disk space but I still wanted to get a feel for this data so I
1:00:00
wanted to get a feel for this data so I
1:00:00
wanted to get a feel for this data so I wrote a little python script that did a
1:00:02
wrote a little python script that did a
1:00:02
wrote a little python script that did a head request against each URL to get
1:00:04
head request against each URL to get
1:00:04
head request against each URL to get back just the length of that file and I
1:00:07
back just the length of that file and I
1:00:07
back just the length of that file and I turned that into a Json
1:00:09
turned that into a Json
1:00:09
turned that into a Json um Json array so here we go this is
1:00:10
um Json array so here we go this is
1:00:10
um Json array so here we go this is saying that file there is this big and
1:00:12
saying that file there is this big and
1:00:12
saying that file there is this big and I've got megabytes and gigabytes as well
1:00:14
I've got megabytes and gigabytes as well
1:00:14
I've got megabytes and gigabytes as well and it's top folders of these very very
1:00:16
and it's top folders of these very very
1:00:16
and it's top folders of these very very simple and what I can do now is I can
1:00:19
simple and what I can do now is I can
1:00:19
simple and what I can do now is I can click load Json and jset like data set
1:00:21
click load Json and jset like data set
1:00:21
click load Json and jset like data set light paste that URL in and it'll fire
1:00:24
light paste that URL in and it'll fire
1:00:24
light paste that URL in and it'll fire it'll import that Json document and turn
1:00:27
it'll import that Json document and turn
1:00:27
it'll import that Json document and turn it into a table so now I can start
1:00:29
it into a table so now I can start
1:00:29
it into a table so now I can start answering questions like what are the
1:00:30
answering questions like what are the
1:00:30
answering questions like what are the top folders it turns out that Wikipedia
1:00:33
top folders it turns out that Wikipedia
1:00:33
top folders it turns out that Wikipedia there's only one file
1:00:34
there's only one file
1:00:34
there's only one file um I think if I sort by size I can see
1:00:37
um I think if I sort by size I can see
1:00:37
um I think if I sort by size I can see that Wikipedia is the largest file they
1:00:39
that Wikipedia is the largest file they
1:00:39
that Wikipedia is the largest file they have a 111 gigabyte file you can
1:00:41
have a 111 gigabyte file you can
1:00:41
have a 111 gigabyte file you can download with all of English Wikipedia
1:00:43
download with all of English Wikipedia
1:00:43
download with all of English Wikipedia in
1:00:44
in
1:00:44
in um but this project here took me five
1:00:46
um but this project here took me five
1:00:46
um but this project here took me five minutes to knock together
1:00:48
minutes to knock together
1:00:48
minutes to knock together um I've got a article about it on my
1:00:49
um I've got a article about it on my
1:00:49
um I've got a article about it on my blog I didn't I literally did the whole
1:00:52
blog I didn't I literally did the whole
1:00:52
blog I didn't I literally did the whole thing copied and paste it into a gist
1:00:54
thing copied and paste it into a gist
1:00:54
thing copied and paste it into a gist assembled this and then started sending
1:00:56
assembled this and then started sending
1:00:56
assembled this and then started sending people this link so you know running a
1:00:59
people this link so you know running a
1:00:59
people this link so you know running a database in webassembly it's actually a
1:01:01
database in webassembly it's actually a
1:01:01
database in webassembly it's actually a really interesting trick
1:01:02
really interesting trick
1:01:02
really interesting trick excitingly you can do this with postgres
1:01:05
excitingly you can do this with postgres
1:01:05
excitingly you can do this with postgres as well this right here is
1:01:08
as well this right here is
1:01:08
as well this right here is um a project but crunchy data built to
1:01:11
um a project but crunchy data built to
1:01:11
um a project but crunchy data built to provide an interactive
1:01:13
provide an interactive
1:01:13
provide an interactive um an interactive uh tutorial for
1:01:16
um an interactive uh tutorial for
1:01:16
um an interactive uh tutorial for learning SQL and this is a full postgres
1:01:18
learning SQL and this is a full postgres
1:01:18
learning SQL and this is a full postgres running in your browser this was
1:01:20
running in your browser this was
1:01:20
running in your browser this was actually inspired by my work on data set
1:01:22
actually inspired by my work on data set
1:01:22
actually inspired by my work on data set light there's one catch it's a 50
1:01:24
light there's one catch it's a 50
1:01:24
light there's one catch it's a 50 megabyte download to get this working
1:01:25
megabyte download to get this working
1:01:25
megabyte download to get this working because they ended up having to do a
1:01:27
because they ended up having to do a
1:01:27
because they ended up having to do a full virtualized Linux operating system
1:01:29
full virtualized Linux operating system
1:01:30
full virtualized Linux operating system with postgres running on top in order in
1:01:32
with postgres running on top in order in
1:01:32
with postgres running on top in order in order to build this I'm very confident
1:01:34
order to build this I'm very confident
1:01:34
order to build this I'm very confident that with a lot of extra work you could
1:01:36
that with a lot of extra work you could
1:01:36
that with a lot of extra work you could shrink this down and just run postcards
1:01:38
shrink this down and just run postcards
1:01:38
shrink this down and just run postcards itself
1:01:39
itself
1:01:39
itself and that brings me to the my sort of
1:01:43
and that brings me to the my sort of
1:01:43
and that brings me to the my sort of call where are we
1:01:45
call where are we
1:01:45
call where are we oh could we switch back to the slides
1:01:50
um
1:01:51
um
1:01:51
um and so the thing that excites the thing
1:01:54
and so the thing that excites the thing
1:01:54
and so the thing that excites the thing the reason I use sqlite for so many of
1:01:56
the reason I use sqlite for so many of
1:01:56
the reason I use sqlite for so many of my projects is that sqlite is a
1:01:58
my projects is that sqlite is a
1:01:58
my projects is that sqlite is a genuinely serverless browser uh database
1:02:01
genuinely serverless browser uh database
1:02:01
genuinely serverless browser uh database it doesn't need a server and this is the
1:02:02
it doesn't need a server and this is the
1:02:03
it doesn't need a server and this is the original meaning of serverless before it
1:02:04
original meaning of serverless before it
1:02:04
original meaning of serverless before it meant whatever it means today in the
1:02:06
meant whatever it means today in the
1:02:06
meant whatever it means today in the cloud there is no service sqlite it's a
1:02:08
cloud there is no service sqlite it's a
1:02:08
cloud there is no service sqlite it's a file on disk it's a C library that opens
1:02:10
file on disk it's a C library that opens
1:02:10
file on disk it's a C library that opens that file everything is is done without
1:02:12
that file everything is is done without
1:02:12
that file everything is is done without network connections and that's really
1:02:15
network connections and that's really
1:02:15
network connections and that's really powerful and useful and something I'm
1:02:17
powerful and useful and something I'm
1:02:17
powerful and useful and something I'm seeing more of these days as well duckdb
1:02:19
seeing more of these days as well duckdb
1:02:19
seeing more of these days as well duckdb has become popular over the past few
1:02:21
has become popular over the past few
1:02:21
has become popular over the past few years and it's exactly the same idea as
1:02:23
years and it's exactly the same idea as
1:02:23
years and it's exactly the same idea as sqlite it's a library you can install
1:02:25
sqlite it's a library you can install
1:02:25
sqlite it's a library you can install and then you can do it it gives you
1:02:27
and then you can do it it gives you
1:02:27
and then you can do it it gives you column there sort of analytical
1:02:30
column there sort of analytical
1:02:30
column there sort of analytical um uh database so like sqlite that
1:02:32
um uh database so like sqlite that
1:02:32
um uh database so like sqlite that optimize those analytical queries and
1:02:34
optimize those analytical queries and
1:02:34
optimize those analytical queries and it's really cool it's a very exciting
1:02:35
it's really cool it's a very exciting
1:02:35
it's really cool it's a very exciting piece of software my question is could
1:02:38
piece of software my question is could
1:02:38
piece of software my question is could postgres do the same thing what if there
1:02:41
postgres do the same thing what if there
1:02:41
postgres do the same thing what if there was a a sort of a bundled version of
1:02:44
was a a sort of a bundled version of
1:02:44
was a a sort of a bundled version of postgres that existed without the
1:02:46
postgres that existed without the
1:02:46
postgres that existed without the network stack without the sort of server
1:02:48
network stack without the sort of server
1:02:48
network stack without the sort of server side of it but it gave you access to the
1:02:51
side of it but it gave you access to the
1:02:51
side of it but it gave you access to the corpus where's some pork put core Plus
1:02:54
corpus where's some pork put core Plus
1:02:54
corpus where's some pork put core Plus squares SQL and data structures and so
1:02:57
squares SQL and data structures and so
1:02:57
squares SQL and data structures and so forth in a way that meant that me as a
1:02:59
forth in a way that meant that me as a
1:02:59
forth in a way that meant that me as a python programmer could pin install it I
1:03:01
python programmer could pin install it I
1:03:01
python programmer could pin install it I would love to be able to say pip install
1:03:03
would love to be able to say pip install
1:03:03
would love to be able to say pip install postgres lib or npm install or whatever
1:03:05
postgres lib or npm install or whatever
1:03:05
postgres lib or npm install or whatever importance of my python code and then
1:03:07
importance of my python code and then
1:03:07
importance of my python code and then use it like I do sqlite induct DB today
1:03:10
use it like I do sqlite induct DB today
1:03:10
use it like I do sqlite induct DB today this would give me the best of all
1:03:12
this would give me the best of all
1:03:12
this would give me the best of all worlds it would give me the feature I
1:03:14
worlds it would give me the feature I
1:03:14
worlds it would give me the feature I care most about from sqlite but it would
1:03:16
care most about from sqlite but it would
1:03:16
care most about from sqlite but it would also give me postgres is far richer and
1:03:19
also give me postgres is far richer and
1:03:19
also give me postgres is far richer and far more feature complete version of SQL
1:03:21
far more feature complete version of SQL
1:03:21
far more feature complete version of SQL and it would give me access to post GIS
1:03:23
and it would give me access to post GIS
1:03:23
and it would give me access to post GIS and extensions like that as well so this
1:03:25
and extensions like that as well so this
1:03:25
and extensions like that as well so this is kind of my dream I think this is if
1:03:27
is kind of my dream I think this is if
1:03:27
is kind of my dream I think this is if this happened I would drop SQL like in a
1:03:30
this happened I would drop SQL like in a
1:03:30
this happened I would drop SQL like in a heartbeat and I would be able to build
1:03:32
heartbeat and I would be able to build
1:03:32
heartbeat and I would be able to build all of my stuff and post squares again
1:03:33
all of my stuff and post squares again
1:03:33
all of my stuff and post squares again so the idea is to take away
1:03:35
so the idea is to take away
1:03:35
so the idea is to take away um small data is really cool it deserves
1:03:37
um small data is really cool it deserves
1:03:37
um small data is really cool it deserves more tooling it's worth thinking about
1:03:39
more tooling it's worth thinking about
1:03:39
more tooling it's worth thinking about how you can build tools for the small
1:03:41
how you can build tools for the small
1:03:41
how you can build tools for the small end as well as for the big end of the
1:03:43
end as well as for the big end of the
1:03:43
end as well as for the big end of the scale
1:03:43
scale
1:03:43
scale the baked data pattern is a way of
1:03:46
the baked data pattern is a way of
1:03:46
the baked data pattern is a way of scaling read-only data both up and down
1:03:49
scaling read-only data both up and down
1:03:49
scaling read-only data both up and down which I think is really powerful and
1:03:51
which I think is really powerful and
1:03:51
which I think is really powerful and gives you a lot of options for for
1:03:52
gives you a lot of options for for
1:03:52
gives you a lot of options for for interesting deployment strategies read
1:03:54
interesting deployment strategies read
1:03:54
interesting deployment strategies read only SQL apis are a great idea forget
1:03:56
only SQL apis are a great idea forget
1:03:56
only SQL apis are a great idea forget about signal injection it's fine just
1:03:59
about signal injection it's fine just
1:03:59
about signal injection it's fine just put if it's public data let people SQL
1:04:01
put if it's public data let people SQL
1:04:01
put if it's public data let people SQL inject all the all that they like and
1:04:03
inject all the all that they like and
1:04:03
inject all the all that they like and yeah you can run databases in the
1:04:05
yeah you can run databases in the
1:04:05
yeah you can run databases in the browser now I never thought that would
1:04:06
browser now I never thought that would
1:04:06
browser now I never thought that would I'd never thought I'd see the day and if
1:04:08
I'd never thought I'd see the day and if
1:04:08
I'd never thought I'd see the day and if you're looking for a very complicated
1:04:10
you're looking for a very complicated
1:04:10
you're looking for a very complicated side project please build me postgres as
1:04:12
side project please build me postgres as
1:04:12
side project please build me postgres as a library I would really appreciate that
1:04:14
a library I would really appreciate that
1:04:14
a library I would really appreciate that and yeah that's um that's that's all
1:04:16
and yeah that's um that's that's all
1:04:16
and yeah that's um that's that's all I've got for you today I will be on
1:04:18
I've got for you today I will be on
1:04:18
I've got for you today I will be on Discord for the rest of the conference
1:04:20
Discord for the rest of the conference
1:04:20
Discord for the rest of the conference and very keen on talking about this and
1:04:21
and very keen on talking about this and
1:04:21
and very keen on talking about this and yeah I've got time for questions
1:04:24
yeah I've got time for questions
1:04:24
yeah I've got time for questions awesome that there's a number of
1:04:27
awesome that there's a number of
1:04:27
awesome that there's a number of different things I loved about that um
1:04:29
different things I loved about that um
1:04:29
different things I loved about that um somebody mentioned in the Discord about
1:04:30
somebody mentioned in the Discord about
1:04:30
somebody mentioned in the Discord about uh the original start with looking at
1:04:32
uh the original start with looking at
1:04:32
uh the original start with looking at data set and how it really opens things
1:04:35
data set and how it really opens things
1:04:35
data set and how it really opens things up to you know the idea of like let's go
1:04:37
up to you know the idea of like let's go
1:04:37
up to you know the idea of like let's go explore the data and not like hear some
1:04:39
explore the data and not like hear some
1:04:39
explore the data and not like hear some pre-compiled thing that you should look
1:04:41
pre-compiled thing that you should look
1:04:41
pre-compiled thing that you should look at in this data it really is about
1:04:42
at in this data it really is about
1:04:42
at in this data it really is about exploring it which I think is really
1:04:43
exploring it which I think is really
1:04:43
exploring it which I think is really fantastic uh so that that was pretty
1:04:46
fantastic uh so that that was pretty
1:04:46
fantastic uh so that that was pretty interesting uh and I and now I feel like
1:04:48
interesting uh and I and now I feel like
1:04:48
interesting uh and I and now I feel like there's a bit of a call to Arms I sort
1:04:50
there's a bit of a call to Arms I sort
1:04:50
there's a bit of a call to Arms I sort of wanted to ask a little bit about how
1:04:51
of wanted to ask a little bit about how
1:04:51
of wanted to ask a little bit about how you integrate this with postgres how
1:04:53
you integrate this with postgres how
1:04:53
you integrate this with postgres how much have you looked at that and I see
1:04:55
much have you looked at that and I see
1:04:55
much have you looked at that and I see that you've kind of got some called arms
1:04:57
that you've kind of got some called arms
1:04:57
that you've kind of got some called arms in there on like well here's what you
1:04:58
in there on like well here's what you
1:04:58
in there on like well here's what you could do with more postgres in there
1:05:01
could do with more postgres in there
1:05:01
could do with more postgres in there um so I've actually built that a little
1:05:03
um so I've actually built that a little
1:05:03
um so I've actually built that a little bit
1:05:04
bit
1:05:04
bit um last year A couple of years ago I
1:05:05
um last year A couple of years ago I
1:05:05
um last year A couple of years ago I built this thing called Django SQL
1:05:07
built this thing called Django SQL
1:05:07
built this thing called Django SQL dashboard where the idea because I had a
1:05:08
dashboard where the idea because I had a
1:05:08
dashboard where the idea because I had a Django app that I wanted to do data set
1:05:11
Django app that I wanted to do data set
1:05:11
Django app that I wanted to do data set like things against and so this is a
1:05:13
like things against and so this is a
1:05:13
like things against and so this is a little Django application I built which
1:05:15
little Django application I built which
1:05:15
little Django application I built which works against the read-only postgres
1:05:17
works against the read-only postgres
1:05:17
works against the read-only postgres thing and gives you some of that data
1:05:18
thing and gives you some of that data
1:05:18
thing and gives you some of that data set functionality the longer term I
1:05:21
set functionality the longer term I
1:05:21
set functionality the longer term I would love data set to have pluggable
1:05:22
would love data set to have pluggable
1:05:22
would love data set to have pluggable back ends if dataset could talk directly
1:05:24
back ends if dataset could talk directly
1:05:24
back ends if dataset could talk directly to post squares I think that would be
1:05:25
to post squares I think that would be
1:05:25
to post squares I think that would be phenomenal and I've shied away from
1:05:27
phenomenal and I've shied away from
1:05:27
phenomenal and I've shied away from doing that because then I have to build
1:05:29
doing that because then I have to build
1:05:29
doing that because then I have to build every feature for sqlite and for
1:05:30
every feature for sqlite and for
1:05:30
every feature for sqlite and for postgres but I'm getting the I'm
1:05:32
postgres but I'm getting the I'm
1:05:32
postgres but I'm getting the I'm beginning to grow a hunch that that's
1:05:33
beginning to grow a hunch that that's
1:05:33
beginning to grow a hunch that that's not as difficult as I thought it would
1:05:35
not as difficult as I thought it would
1:05:35
not as difficult as I thought it would be the the SQL dialects are very similar
1:05:37
be the the SQL dialects are very similar
1:05:37
be the the SQL dialects are very similar between the two so it might happen yeah
1:05:39
between the two so it might happen yeah
1:05:39
between the two so it might happen yeah absolutely I've had the opportunity to
1:05:41
absolutely I've had the opportunity to
1:05:41
absolutely I've had the opportunity to talk to Richard uh hip from sqlite a
1:05:43
talk to Richard uh hip from sqlite a
1:05:43
talk to Richard uh hip from sqlite a number of times and he's always stated
1:05:45
number of times and he's always stated
1:05:45
number of times and he's always stated that he works at the postgres you know
1:05:47
that he works at the postgres you know
1:05:47
that he works at the postgres you know syntax and language as a sort of a first
1:05:49
syntax and language as a sort of a first
1:05:49
syntax and language as a sort of a first implementation for him to base what he
1:05:51
implementation for him to base what he
1:05:51
implementation for him to base what he does so I think that that would actually
1:05:53
does so I think that that would actually
1:05:53
does so I think that that would actually be pretty pretty easy to to make that
1:05:55
be pretty pretty easy to to make that
1:05:55
be pretty pretty easy to to make that all put together so
1:05:57
all put together so
1:05:57
all put together so sounds awesome what I like about your
1:05:59
sounds awesome what I like about your
1:05:59
sounds awesome what I like about your talk Simon is just that you're you're
1:06:01
talk Simon is just that you're you're
1:06:01
talk Simon is just that you're you're looking at postgres from right next door
1:06:03
looking at postgres from right next door
1:06:03
looking at postgres from right next door right you're not actively spending all
1:06:06
right you're not actively spending all
1:06:06
right you're not actively spending all of your days working on postgres
1:06:09
of your days working on postgres
1:06:09
of your days working on postgres um but you are solving similar sorts of
1:06:11
um but you are solving similar sorts of
1:06:11
um but you are solving similar sorts of problems with other tools and I think
1:06:14
problems with other tools and I think
1:06:14
problems with other tools and I think you and I talked about this as a
1:06:15
you and I talked about this as a
1:06:15
you and I talked about this as a postgres adjacent talk absolutely when
1:06:18
postgres adjacent talk absolutely when
1:06:18
postgres adjacent talk absolutely when we were first conceptualizing it and and
1:06:19
we were first conceptualizing it and and
1:06:19
we were first conceptualizing it and and I think it's great because you see
1:06:21
I think it's great because you see
1:06:21
I think it's great because you see things in a way that's different from
1:06:23
things in a way that's different from
1:06:23
things in a way that's different from those of us who are like in this circle
1:06:26
those of us who are like in this circle
1:06:26
those of us who are like in this circle all the time
1:06:28
all the time
1:06:28
all the time so I I would actually love to chat more
1:06:31
so I I would actually love to chat more
1:06:31
so I I would actually love to chat more I think we are Discord description yeah
1:06:34
I think we are Discord description yeah
1:06:34
I think we are Discord description yeah so there are questions on the Discord uh
1:06:36
so there are questions on the Discord uh
1:06:36
so there are questions on the Discord uh I will be checking that out as well
1:06:38
I will be checking that out as well
1:06:38
I will be checking that out as well um but I think we actually need to move
1:06:40
um but I think we actually need to move
1:06:40
um but I think we actually need to move on to our next speaker Simon thank you
1:06:42
on to our next speaker Simon thank you
1:06:42
on to our next speaker Simon thank you so much for being here absolutely 100
1:06:44
so much for being here absolutely 100
1:06:44
so much for being here absolutely 100 that was awesome thanks very much fat
1:06:46
that was awesome thanks very much fat
1:06:46
that was awesome thanks very much fat man yeah I'll be on Discord in just a
1:06:48
man yeah I'll be on Discord in just a
1:06:48
man yeah I'll be on Discord in just a few seconds and thank you for your path
1:06:49
few seconds and thank you for your path
1:06:49
few seconds and thank you for your path to cytuscon episode one uh which I hope
1:06:52
to cytuscon episode one uh which I hope
1:06:52
to cytuscon episode one uh which I hope people catch
1:06:53
people catch
1:06:53
people catch um really really interesting
1:06:54
um really really interesting
1:06:54
um really really interesting conversation on working in public too
1:06:56
conversation on working in public too
1:06:56
conversation on working in public too cool have a great day
1:07:00
so our next speaker is going to be yelta
1:07:04
so our next speaker is going to be yelta
1:07:04
so our next speaker is going to be yelta fenima and he is speaking uh ironically
1:07:08
fenima and he is speaking uh ironically
1:07:08
fenima and he is speaking uh ironically uh because of the the little niche
1:07:11
uh because of the the little niche
1:07:11
uh because of the the little niche Museum in Simon's talk I love that like
1:07:13
Museum in Simon's talk I love that like
1:07:13
Museum in Simon's talk I love that like you know here's the apology from AI for
1:07:15
you know here's the apology from AI for
1:07:15
you know here's the apology from AI for destroying the world so of course up
1:07:17
destroying the world so of course up
1:07:17
destroying the world so of course up next we need to and this is this is not
1:07:19
next we need to and this is this is not
1:07:19
next we need to and this is this is not planned we need to have a talk about
1:07:20
planned we need to have a talk about
1:07:20
planned we need to have a talk about Chad gbt and postgres and rust and all
1:07:23
Chad gbt and postgres and rust and all
1:07:23
Chad gbt and postgres and rust and all that so uh that's fantastic how you
1:07:25
that so uh that's fantastic how you
1:07:25
that so uh that's fantastic how you doing Yota I'm doing well yeah so I'm
1:07:28
doing Yota I'm doing well yeah so I'm
1:07:28
doing Yota I'm doing well yeah so I'm very excited you're joining us from the
1:07:30
very excited you're joining us from the
1:07:30
very excited you're joining us from the Netherlands I am yes
1:07:32
Netherlands I am yes
1:07:32
Netherlands I am yes excellent uh so just for those who have
1:07:35
excellent uh so just for those who have
1:07:35
excellent uh so just for those who have not met yelta before he's a senior
1:07:37
not met yelta before he's a senior
1:07:37
not met yelta before he's a senior software engineer at Microsoft uh you
1:07:39
software engineer at Microsoft uh you
1:07:39
software engineer at Microsoft uh you work on cytus uh and on postgres and I
1:07:43
work on cytus uh and on postgres and I
1:07:43
work on cytus uh and on postgres and I believe you're a maintainer on PG
1:07:44
believe you're a maintainer on PG
1:07:44
believe you're a maintainer on PG bouncer yes all of those yeah so so we
1:07:47
bouncer yes all of those yeah so so we
1:07:47
bouncer yes all of those yeah so so we all owe you uh a few beers I imagine uh
1:07:50
all owe you uh a few beers I imagine uh
1:07:50
all owe you uh a few beers I imagine uh for that for work across the board so
1:07:52
for that for work across the board so
1:07:52
for that for work across the board so excellent
1:07:54
excellent
1:07:54
excellent or coffee coffee works too or a coffee
1:07:56
or coffee coffee works too or a coffee
1:07:56
or coffee coffee works too or a coffee yeah
1:07:57
yeah
1:07:57
yeah okay
1:07:58
okay
1:07:58
okay um well yelta take us away
1:08:01
um well yelta take us away
1:08:01
um well yelta take us away all right
1:08:03
all right
1:08:03
all right so I'm gonna talk a bit about this whole
1:08:07
so I'm gonna talk a bit about this whole
1:08:07
so I'm gonna talk a bit about this whole thing with AI that if you've been on the
1:08:10
thing with AI that if you've been on the
1:08:10
thing with AI that if you've been on the internet for the past few months uh you
1:08:13
internet for the past few months uh you
1:08:13
internet for the past few months uh you saw happening that's sort of taking over
1:08:15
saw happening that's sort of taking over
1:08:15
saw happening that's sort of taking over the world everyone's like oh what's it
1:08:16
the world everyone's like oh what's it
1:08:16
the world everyone's like oh what's it going what you cannot do everything with
1:08:18
going what you cannot do everything with
1:08:18
going what you cannot do everything with that yeah okay great pictures you don't
1:08:20
that yeah okay great pictures you don't
1:08:20
that yeah okay great pictures you don't need to paint anymore uh you can
1:08:24
need to paint anymore uh you can
1:08:24
need to paint anymore uh you can do anything you don't need to Google you
1:08:26
do anything you don't need to Google you
1:08:27
do anything you don't need to Google you can just ask Ai and it knows
1:08:30
can just ask Ai and it knows
1:08:30
can just ask Ai and it knows so I'm I'm here a bit for how can you
1:08:34
so I'm I'm here a bit for how can you
1:08:34
so I'm I'm here a bit for how can you use this AI for postgres what what can
1:08:38
use this AI for postgres what what can
1:08:38
use this AI for postgres what what can you do to integrate it with postgres and
1:08:41
you do to integrate it with postgres and
1:08:41
you do to integrate it with postgres and how can you use it in useful ways
1:08:44
how can you use it in useful ways
1:08:44
how can you use it in useful ways and uh yeah
1:08:46
and uh yeah
1:08:46
and uh yeah that's uh that's gonna be my talk
1:08:48
that's uh that's gonna be my talk
1:08:48
that's uh that's gonna be my talk so the reason I'm doing this talk is
1:08:51
so the reason I'm doing this talk is
1:08:51
so the reason I'm doing this talk is because I have one secret problem I've
1:08:53
because I have one secret problem I've
1:08:53
because I have one secret problem I've been working on postgres for uh with
1:08:56
been working on postgres for uh with
1:08:56
been working on postgres for uh with postgres on postgres for quite a few
1:08:58
postgres on postgres for quite a few
1:08:58
postgres on postgres for quite a few years now
1:08:59
years now
1:08:59
years now uh
1:09:01
uh
1:09:01
uh and I still have it's very difficult
1:09:03
and I still have it's very difficult
1:09:03
and I still have it's very difficult time to write working postgres queries
1:09:06
time to write working postgres queries
1:09:06
time to write working postgres queries I don't know what it is it's just I
1:09:08
I don't know what it is it's just I
1:09:08
I don't know what it is it's just I forget a comma I swap around order buy
1:09:11
forget a comma I swap around order buy
1:09:11
forget a comma I swap around order buy and group buy and then it's like oh it's
1:09:13
and group buy and then it's like oh it's
1:09:13
and group buy and then it's like oh it's the wrong order it doesn't work at all
1:09:14
the wrong order it doesn't work at all
1:09:14
the wrong order it doesn't work at all uh I forget how many parentheses I need
1:09:17
uh I forget how many parentheses I need
1:09:17
uh I forget how many parentheses I need for your values statements all of those
1:09:19
for your values statements all of those
1:09:19
for your values statements all of those things can happen and
1:09:21
things can happen and
1:09:21
things can happen and I don't know it's I my queries never
1:09:23
I don't know it's I my queries never
1:09:23
I don't know it's I my queries never work first try and I don't think I'm
1:09:25
work first try and I don't think I'm
1:09:25
work first try and I don't think I'm alone at least I hope
1:09:28
alone at least I hope
1:09:28
alone at least I hope so
1:09:29
so
1:09:29
so in the past it was always just sort of a
1:09:31
in the past it was always just sort of a
1:09:31
in the past it was always just sort of a given it's like okay I'll just write the
1:09:32
given it's like okay I'll just write the
1:09:32
given it's like okay I'll just write the query again but now with the AI hacking
1:09:35
query again but now with the AI hacking
1:09:35
query again but now with the AI hacking I was like oh maybe I can ask jet GPT to
1:09:37
I was like oh maybe I can ask jet GPT to
1:09:37
I was like oh maybe I can ask jet GPT to help me
1:09:38
help me
1:09:38
help me so that's that's what I did and chat GPT
1:09:41
so that's that's what I did and chat GPT
1:09:41
so that's that's what I did and chat GPT was like yes of course I can I can help
1:09:44
was like yes of course I can I can help
1:09:44
was like yes of course I can I can help you what uh what do you need
1:09:46
you what uh what do you need
1:09:46
you what uh what do you need and I was like okay well uh what do you
1:09:48
and I was like okay well uh what do you
1:09:48
and I was like okay well uh what do you need to know for me for I mean what do
1:09:51
need to know for me for I mean what do
1:09:51
need to know for me for I mean what do you need me to tell you for me for you
1:09:53
you need me to tell you for me for you
1:09:53
you need me to tell you for me for you to help me
1:09:53
to help me
1:09:54
to help me and it wanted to know some tables and
1:09:56
and it wanted to know some tables and
1:09:56
and it wanted to know some tables and column names and like the things I
1:09:57
column names and like the things I
1:09:57
column names and like the things I wanted to do of course
1:09:59
wanted to do of course
1:09:59
wanted to do of course so that's and then I thought that that's
1:10:02
so that's and then I thought that that's
1:10:02
so that's and then I thought that that's that's a bit annoying then I have to
1:10:04
that's a bit annoying then I have to
1:10:04
that's a bit annoying then I have to copy paste my schema from my database or
1:10:07
copy paste my schema from my database or
1:10:07
copy paste my schema from my database or from some file to
1:10:08
from some file to
1:10:08
from some file to into captivity so notes all the tables
1:10:10
into captivity so notes all the tables
1:10:10
into captivity so notes all the tables or the columns I have
1:10:13
or the columns I have
1:10:13
or the columns I have and that's another problem I have a
1:10:15
and that's another problem I have a
1:10:15
and that's another problem I have a little not so secret is that I don't
1:10:17
little not so secret is that I don't
1:10:17
little not so secret is that I don't like doing boring things and copy
1:10:19
like doing boring things and copy
1:10:19
like doing boring things and copy pressing schemas increase I mean to me
1:10:21
pressing schemas increase I mean to me
1:10:21
pressing schemas increase I mean to me that sounds extremely boring uh so I
1:10:25
that sounds extremely boring uh so I
1:10:25
that sounds extremely boring uh so I thought how can I how can I avoid this
1:10:28
thought how can I how can I avoid this
1:10:28
thought how can I how can I avoid this boring stuff
1:10:29
boring stuff
1:10:29
boring stuff and
1:10:30
and
1:10:30
and because password has all this
1:10:32
because password has all this
1:10:32
because password has all this information maybe I can get postgres to
1:10:35
information maybe I can get postgres to
1:10:35
information maybe I can get postgres to to communicate Rich activity for me
1:10:38
to communicate Rich activity for me
1:10:38
to communicate Rich activity for me instead of doing it myself directly
1:10:41
instead of doing it myself directly
1:10:41
instead of doing it myself directly and that's
1:10:43
and that's
1:10:43
and that's that's I mean I would that would need
1:10:46
that's I mean I would that would need
1:10:46
that's I mean I would that would need some functionality in postgres to be
1:10:48
some functionality in postgres to be
1:10:48
some functionality in postgres to be able to do that
1:10:50
able to do that
1:10:50
able to do that and that's actually quite possible
1:10:52
and that's actually quite possible
1:10:52
and that's actually quite possible because postgres has extensions it's one
1:10:55
because postgres has extensions it's one
1:10:55
because postgres has extensions it's one of the
1:10:56
of the
1:10:56
of the thing it's one of the things that make
1:10:57
thing it's one of the things that make
1:10:57
thing it's one of the things that make Costco's quite special with with respect
1:11:00
Costco's quite special with with respect
1:11:00
Costco's quite special with with respect to other relational databases it's
1:11:02
to other relational databases it's
1:11:02
to other relational databases it's actually in the original paper from bus
1:11:04
actually in the original paper from bus
1:11:04
actually in the original paper from bus for postgres like the abstract
1:11:07
for postgres like the abstract
1:11:07
for postgres like the abstract uh that postgres is it's one of its
1:11:10
uh that postgres is it's one of its
1:11:10
uh that postgres is it's one of its goals is to be extensible
1:11:12
goals is to be extensible
1:11:12
goals is to be extensible and the way it does that is that you can
1:11:15
and the way it does that is that you can
1:11:15
and the way it does that is that you can Define your own types your own functions
1:11:16
Define your own types your own functions
1:11:16
Define your own types your own functions and how you do that is by having two
1:11:19
and how you do that is by having two
1:11:19
and how you do that is by having two parts you have a shared Library which
1:11:21
parts you have a shared Library which
1:11:21
parts you have a shared Library which contains all the native code
1:11:24
contains all the native code
1:11:24
contains all the native code um
1:11:25
um
1:11:25
um so it can that can do anything basically
1:11:28
so it can that can do anything basically
1:11:28
so it can that can do anything basically it's like any program running on your
1:11:29
it's like any program running on your
1:11:29
it's like any program running on your computer can do can make web requests
1:11:31
computer can do can make web requests
1:11:31
computer can do can make web requests can do you can do whatever
1:11:33
can do you can do whatever
1:11:33
can do you can do whatever uh and then sort of to to call those
1:11:36
uh and then sort of to to call those
1:11:36
uh and then sort of to to call those Nate to call that native code from the
1:11:39
Nate to call that native code from the
1:11:39
Nate to call that native code from the SQL function
1:11:40
SQL function
1:11:40
SQL function uh from a SQL query you define some SQL
1:11:44
uh from a SQL query you define some SQL
1:11:44
uh from a SQL query you define some SQL functions and those SQL functions are
1:11:46
functions and those SQL functions are
1:11:46
functions and those SQL functions are very simple they're like ah this is the
1:11:47
very simple they're like ah this is the
1:11:47
very simple they're like ah this is the name of the SQL function this is the
1:11:49
name of the SQL function this is the
1:11:49
name of the SQL function this is the things that returns and arguments it
1:11:50
things that returns and arguments it
1:11:50
things that returns and arguments it takes and this is the Nate the name of
1:11:52
takes and this is the Nate the name of
1:11:52
takes and this is the Nate the name of the Native function that it will
1:11:54
the Native function that it will
1:11:54
the Native function that it will actually call internally
1:11:57
actually call internally
1:11:57
actually call internally and scientist is built like that lots of
1:11:59
and scientist is built like that lots of
1:11:59
and scientist is built like that lots of other things are built like that so I
1:12:01
other things are built like that so I
1:12:01
other things are built like that so I thought oh let's write a GPT
1:12:03
thought oh let's write a GPT
1:12:03
thought oh let's write a GPT postgres extension uh how hard can it be
1:12:07
postgres extension uh how hard can it be
1:12:07
postgres extension uh how hard can it be and uh there's one other thing about me
1:12:09
and uh there's one other thing about me
1:12:09
and uh there's one other thing about me I like rust uh it's like it's like this
1:12:12
I like rust uh it's like it's like this
1:12:12
I like rust uh it's like it's like this fancy new sort of new language uh it's I
1:12:15
fancy new sort of new language uh it's I
1:12:15
fancy new sort of new language uh it's I mean it's not super new anymore but it
1:12:17
mean it's not super new anymore but it
1:12:17
mean it's not super new anymore but it still sort of feels new uh it all the
1:12:20
still sort of feels new uh it all the
1:12:20
still sort of feels new uh it all the all the things about it there are feel
1:12:22
all the things about it there are feel
1:12:22
all the things about it there are feel new it has like everything is an
1:12:24
new it has like everything is an
1:12:24
new it has like everything is an expression so even your if you're else
1:12:26
expression so even your if you're else
1:12:26
expression so even your if you're else they sort of return values and you can
1:12:28
they sort of return values and you can
1:12:28
they sort of return values and you can do better metric which is like it's I
1:12:30
do better metric which is like it's I
1:12:30
do better metric which is like it's I mean it's it's sort of destructuring of
1:12:33
mean it's it's sort of destructuring of
1:12:33
mean it's it's sort of destructuring of of a date of structures and structs and
1:12:36
of a date of structures and structs and
1:12:36
of a date of structures and structs and objects and and sort of a it sort of the
1:12:40
objects and and sort of a it sort of the
1:12:40
objects and and sort of a it sort of the way that it just looks nicely called and
1:12:42
way that it just looks nicely called and
1:12:42
way that it just looks nicely called and it's I mean it's hard to explain if you
1:12:44
it's I mean it's hard to explain if you
1:12:44
it's I mean it's hard to explain if you haven't seen it just go look at it
1:12:47
haven't seen it just go look at it
1:12:47
haven't seen it just go look at it um and the type system is also fancy
1:12:48
um and the type system is also fancy
1:12:48
um and the type system is also fancy it's like Haskell but I can actually
1:12:51
it's like Haskell but I can actually
1:12:51
it's like Haskell but I can actually understand what it does
1:12:54
understand what it does
1:12:54
understand what it does um
1:12:54
um
1:12:54
um and it will give you things like not
1:12:57
and it will give you things like not
1:12:57
and it will give you things like not having no points or exceptions
1:12:59
having no points or exceptions
1:12:59
having no points or exceptions and
1:13:00
and
1:13:00
and on top of that there's also extremely
1:13:02
on top of that there's also extremely
1:13:02
on top of that there's also extremely secure there's no there's no memory
1:13:04
secure there's no there's no memory
1:13:04
secure there's no there's no memory leaks no double threes no data races
1:13:06
leaks no double threes no data races
1:13:06
leaks no double threes no data races none of those things that you're used to
1:13:07
none of those things that you're used to
1:13:07
none of those things that you're used to from from C or super cluster and data
1:13:10
from from C or super cluster and data
1:13:10
from from C or super cluster and data races can happen in pretty much any
1:13:11
races can happen in pretty much any
1:13:11
races can happen in pretty much any language but even though those that
1:13:14
language but even though those that
1:13:14
language but even though those that don't I mean if you're doing sort of the
1:13:16
don't I mean if you're doing sort of the
1:13:16
don't I mean if you're doing sort of the in the correct way that they don't
1:13:18
in the correct way that they don't
1:13:18
in the correct way that they don't happen in Rust and and even with all
1:13:21
happen in Rust and and even with all
1:13:21
happen in Rust and and even with all those security things it's just as fast
1:13:25
those security things it's just as fast
1:13:25
those security things it's just as fast as SQL plus I mean and see it's it
1:13:27
as SQL plus I mean and see it's it
1:13:27
as SQL plus I mean and see it's it although I mean it doesn't really matter
1:13:28
although I mean it doesn't really matter
1:13:28
although I mean it doesn't really matter for our use case because I mean I just
1:13:31
for our use case because I mean I just
1:13:31
for our use case because I mean I just want to talk to chatgpt and probably
1:13:33
want to talk to chatgpt and probably
1:13:33
want to talk to chatgpt and probably chat GPT is the slow thing in this whole
1:13:36
chat GPT is the slow thing in this whole
1:13:36
chat GPT is the slow thing in this whole it's all set up but I mean we can at
1:13:39
it's all set up but I mean we can at
1:13:39
it's all set up but I mean we can at least sense
1:13:40
least sense
1:13:40
least sense questions to share activity very very
1:13:42
questions to share activity very very
1:13:42
questions to share activity very very fast and how long that how long chapter
1:13:44
fast and how long that how long chapter
1:13:44
fast and how long that how long chapter it takes to respond that's uh that's
1:13:46
it takes to respond that's uh that's
1:13:46
it takes to respond that's uh that's another another method
1:13:48
another another method
1:13:48
another another method and finally not unimportant it has a it
1:13:51
and finally not unimportant it has a it
1:13:51
and finally not unimportant it has a it has a package manager and see C2 plus
1:13:53
has a package manager and see C2 plus
1:13:53
has a package manager and see C2 plus they famously don't and managing
1:13:55
they famously don't and managing
1:13:55
they famously don't and managing dependencies it's it's really it's uh
1:13:58
dependencies it's it's really it's uh
1:13:58
dependencies it's it's really it's uh horrible
1:13:59
horrible
1:13:59
horrible um
1:14:00
um
1:14:00
um that's that's kind of the yeah so
1:14:03
that's that's kind of the yeah so
1:14:03
that's that's kind of the yeah so package eventually rest is good
1:14:05
package eventually rest is good
1:14:05
package eventually rest is good it's I mean like any other package
1:14:07
it's I mean like any other package
1:14:07
it's I mean like any other package manager for most languages but cnc4 plus
1:14:09
manager for most languages but cnc4 plus
1:14:09
manager for most languages but cnc4 plus they don't have it
1:14:11
they don't have it
1:14:11
they don't have it and finally sort of bonus thing it has
1:14:13
and finally sort of bonus thing it has
1:14:13
and finally sort of bonus thing it has this really cute little crab as a mascot
1:14:16
this really cute little crab as a mascot
1:14:16
this really cute little crab as a mascot he's called Ferris so that's uh I mean
1:14:19
he's called Ferris so that's uh I mean
1:14:19
he's called Ferris so that's uh I mean bonus points for cute mascots
1:14:23
bonus points for cute mascots
1:14:24
bonus points for cute mascots so I've actually been playing with rust
1:14:26
so I've actually been playing with rust
1:14:26
so I've actually been playing with rust for quite a long time I I have a fairly
1:14:29
for quite a long time I I have a fairly
1:14:29
for quite a long time I I have a fairly popular open source Library called
1:14:31
popular open source Library called
1:14:31
popular open source Library called derive more
1:14:33
derive more
1:14:33
derive more um and it's it's ultimates writing
1:14:35
um and it's it's ultimates writing
1:14:35
um and it's it's ultimates writing boring boilerplating rushed
1:14:37
boring boilerplating rushed
1:14:37
boring boilerplating rushed so you see you can might see some
1:14:40
so you see you can might see some
1:14:40
so you see you can might see some some yeah some things in the things I
1:14:43
some yeah some things in the things I
1:14:43
some yeah some things in the things I work some commonalities of the things I
1:14:45
work some commonalities of the things I
1:14:45
work some commonalities of the things I work on where I'm trying to automate all
1:14:47
work on where I'm trying to automate all
1:14:47
work on where I'm trying to automate all the boring things away that's kind of
1:14:49
the boring things away that's kind of
1:14:49
the boring things away that's kind of what I do
1:14:50
what I do
1:14:50
what I do um and I actually did this seven seven
1:14:52
um and I actually did this seven seven
1:14:52
um and I actually did this seven seven years ago I started with it and it I
1:14:54
years ago I started with it and it I
1:14:54
years ago I started with it and it I mean I mean it grew way more than I
1:14:56
mean I mean it grew way more than I
1:14:56
mean I mean it grew way more than I expected uh because apparently other
1:14:58
expected uh because apparently other
1:14:58
expected uh because apparently other developers also don't like doing boring
1:15:00
developers also don't like doing boring
1:15:00
developers also don't like doing boring things so now it's it's actually used by
1:15:02
things so now it's it's actually used by
1:15:02
things so now it's it's actually used by more than 100
1:15:03
more than 100
1:15:03
more than 100 000
1:15:04
000
1:15:04
000 repositories on GitHub so it's quite I
1:15:07
repositories on GitHub so it's quite I
1:15:07
repositories on GitHub so it's quite I mean I uh that's definitely my most used
1:15:10
mean I uh that's definitely my most used
1:15:10
mean I uh that's definitely my most used library for sure
1:15:13
library for sure
1:15:13
library for sure so I never used rust professionally I
1:15:17
so I never used rust professionally I
1:15:17
so I never used rust professionally I only did this sort of as like I started
1:15:19
only did this sort of as like I started
1:15:19
only did this sort of as like I started in I started doing this at the end of my
1:15:21
in I started doing this at the end of my
1:15:21
in I started doing this at the end of my University and then uh
1:15:23
University and then uh
1:15:23
University and then uh yeah it I sort of did it outside of
1:15:26
yeah it I sort of did it outside of
1:15:26
yeah it I sort of did it outside of outside of work I sort of cooked it a
1:15:28
outside of work I sort of cooked it a
1:15:28
outside of work I sort of cooked it a bit with it so that's something I kind
1:15:30
bit with it so that's something I kind
1:15:30
bit with it so that's something I kind of want to change so how do I combine
1:15:33
of want to change so how do I combine
1:15:33
of want to change so how do I combine postgres and rust and maybe even cite
1:15:35
postgres and rust and maybe even cite
1:15:35
postgres and rust and maybe even cite this because I mean I'm working on
1:15:36
this because I mean I'm working on
1:15:36
this because I mean I'm working on scientists much as much of the time so I
1:15:38
scientists much as much of the time so I
1:15:38
scientists much as much of the time so I could even even sort of two sides it to
1:15:40
could even even sort of two sides it to
1:15:40
could even even sort of two sides it to make it more relatable to work and get
1:15:42
make it more relatable to work and get
1:15:42
make it more relatable to work and get Microsoft to pay you for doing doing
1:15:45
Microsoft to pay you for doing doing
1:15:45
Microsoft to pay you for doing doing stuff on uh with rust
1:15:49
and actually I'm not the only one that
1:15:52
and actually I'm not the only one that
1:15:52
and actually I'm not the only one that that wants to combine postgres and rest
1:15:55
that wants to combine postgres and rest
1:15:55
that wants to combine postgres and rest there's there's this great PGX library
1:15:59
there's there's this great PGX library
1:15:59
there's there's this great PGX library and I just learned today that like in
1:16:02
and I just learned today that like in
1:16:02
and I just learned today that like in the last 24 hours they changed their
1:16:04
the last 24 hours they changed their
1:16:04
the last 24 hours they changed their name to pgrx so the I guess the r stands
1:16:08
name to pgrx so the I guess the r stands
1:16:08
name to pgrx so the I guess the r stands for rust
1:16:09
for rust
1:16:09
for rust um so it's that's I mean so anything
1:16:12
um so it's that's I mean so anything
1:16:12
um so it's that's I mean so anything that says PTX in the next slides it's
1:16:14
that says PTX in the next slides it's
1:16:14
that says PTX in the next slides it's like uh probably it will be pgrx
1:16:18
like uh probably it will be pgrx
1:16:18
like uh probably it will be pgrx tomorrow
1:16:19
tomorrow
1:16:19
tomorrow uh but yeah I couldn't change my slides
1:16:21
uh but yeah I couldn't change my slides
1:16:21
uh but yeah I couldn't change my slides anymore in time
1:16:24
anymore in time
1:16:24
anymore in time but uh yeah this is a rough library to
1:16:27
but uh yeah this is a rough library to
1:16:27
but uh yeah this is a rough library to build postgres extensions with rust and
1:16:29
build postgres extensions with rust and
1:16:29
build postgres extensions with rust and that's kind of just what I want to do
1:16:32
that's kind of just what I want to do
1:16:32
that's kind of just what I want to do and it's actually really easy to use
1:16:33
and it's actually really easy to use
1:16:33
and it's actually really easy to use it's uh it's a few commands you install
1:16:36
it's uh it's a few commands you install
1:16:36
it's uh it's a few commands you install cargo PGA cargo PGX cargo is the rest
1:16:40
cargo PGA cargo PGX cargo is the rest
1:16:40
cargo PGA cargo PGX cargo is the rest package manager and you can install some
1:16:42
package manager and you can install some
1:16:42
package manager and you can install some plugins to it so by by doing by running
1:16:46
plugins to it so by by doing by running
1:16:46
plugins to it so by by doing by running cargo install cargo PGX you can then run
1:16:49
cargo install cargo PGX you can then run
1:16:49
cargo install cargo PGX you can then run cargo PGX in it to sort of set up
1:16:52
cargo PGX in it to sort of set up
1:16:52
cargo PGX in it to sort of set up postgres uh
1:16:54
postgres uh
1:16:54
postgres uh extension tooling
1:16:57
extension tooling
1:16:57
extension tooling and then you can create a new extension
1:16:59
and then you can create a new extension
1:16:59
and then you can create a new extension you could do cargo PGX new my extension
1:17:01
you could do cargo PGX new my extension
1:17:01
you could do cargo PGX new my extension and you go to the directory of the
1:17:02
and you go to the directory of the
1:17:02
and you go to the directory of the extension you can play around with it
1:17:04
extension you can play around with it
1:17:04
extension you can play around with it that's that's in the readme of the of
1:17:06
that's that's in the readme of the of
1:17:06
that's that's in the readme of the of the project
1:17:07
the project
1:17:07
the project and it's
1:17:09
and it's
1:17:09
and it's but that sort of brings us to a next
1:17:11
but that sort of brings us to a next
1:17:11
but that sort of brings us to a next issue
1:17:13
issue
1:17:13
issue I need a name for this cool new GPT
1:17:16
I need a name for this cool new GPT
1:17:16
I need a name for this cool new GPT extension so I mean sort of the most
1:17:17
extension so I mean sort of the most
1:17:17
extension so I mean sort of the most obvious one is tgpt
1:17:20
obvious one is tgpt
1:17:20
obvious one is tgpt but it turns out that I wasn't the only
1:17:22
but it turns out that I wasn't the only
1:17:22
but it turns out that I wasn't the only one with this nice idea of combining
1:17:25
one with this nice idea of combining
1:17:25
one with this nice idea of combining things so someone else sort of did
1:17:27
things so someone else sort of did
1:17:27
things so someone else sort of did exactly what I wanted to do but I
1:17:28
exactly what I wanted to do but I
1:17:28
exactly what I wanted to do but I already served into the top and I mean I
1:17:30
already served into the top and I mean I
1:17:30
already served into the top and I mean I still think it's I already started
1:17:32
still think it's I already started
1:17:32
still think it's I already started working on it also so it's uh it's it
1:17:35
working on it also so it's uh it's it
1:17:35
working on it also so it's uh it's it and it's of course it's different mine
1:17:37
and it's of course it's different mine
1:17:37
and it's of course it's different mine is obviously better
1:17:39
is obviously better
1:17:39
is obviously better um but but yeah that's so but I should
1:17:41
um but but yeah that's so but I should
1:17:41
um but but yeah that's so but I should at least not take the same name that's
1:17:43
at least not take the same name that's
1:17:43
at least not take the same name that's that's just confusing for uh for
1:17:45
that's just confusing for uh for
1:17:45
that's just confusing for uh for everyone so maybe flip it around go for
1:17:48
everyone so maybe flip it around go for
1:17:48
everyone so maybe flip it around go for GPT PG and I mean it's a nice palindrome
1:17:52
GPT PG and I mean it's a nice palindrome
1:17:52
GPT PG and I mean it's a nice palindrome so that's kind of cool but it's it's
1:17:53
so that's kind of cool but it's it's
1:17:53
so that's kind of cool but it's it's kind of hard to pronounce the only
1:17:54
kind of hard to pronounce the only
1:17:54
kind of hard to pronounce the only reason that that role of my tongue so
1:17:56
reason that that role of my tongue so
1:17:56
reason that that role of my tongue so easily because I've practiced it a lot
1:17:59
easily because I've practiced it a lot
1:17:59
easily because I've practiced it a lot um so so I ended up as the final choice
1:18:01
um so so I ended up as the final choice
1:18:01
um so so I ended up as the final choice of PG human it doesn't cook 10gbt but
1:18:04
of PG human it doesn't cook 10gbt but
1:18:04
of PG human it doesn't cook 10gbt but it's it sort of brings the idea to you
1:18:07
it's it sort of brings the idea to you
1:18:07
it's it sort of brings the idea to you like ah we want to humanize postgres so
1:18:09
like ah we want to humanize postgres so
1:18:09
like ah we want to humanize postgres so if you don't
1:18:10
if you don't
1:18:10
if you don't it's not such a machine anymore it
1:18:12
it's not such a machine anymore it
1:18:12
it's not such a machine anymore it becomes becomes a bit more human
1:18:15
becomes becomes a bit more human
1:18:15
becomes becomes a bit more human it understands you better
1:18:18
so I double checked the chat utility if
1:18:20
so I double checked the chat utility if
1:18:20
so I double checked the chat utility if it thought it was a good name for the
1:18:23
it thought it was a good name for the
1:18:23
it thought it was a good name for the extension and it was like yes oh yeah
1:18:25
extension and it was like yes oh yeah
1:18:25
extension and it was like yes oh yeah it's a great name short catchy easy to
1:18:27
it's a great name short catchy easy to
1:18:27
it's a great name short catchy easy to remember
1:18:28
remember
1:18:28
remember so I was like okay well if GPT
1:18:31
so I was like okay well if GPT
1:18:31
so I was like okay well if GPT agrees I will use this name
1:18:34
agrees I will use this name
1:18:34
agrees I will use this name so I continued I changed my extension to
1:18:37
so I continued I changed my extension to
1:18:37
so I continued I changed my extension to PG human so now we have a directory and
1:18:41
PG human so now we have a directory and
1:18:41
PG human so now we have a directory and then in the directory you run cargo PGX
1:18:44
then in the directory you run cargo PGX
1:18:44
then in the directory you run cargo PGX run and you create the extension inside
1:18:47
run and you create the extension inside
1:18:47
run and you create the extension inside the postgresh shell which is Cargo PDX
1:18:49
the postgresh shell which is Cargo PDX
1:18:49
the postgresh shell which is Cargo PDX run actually spawns the possible SQL
1:18:51
run actually spawns the possible SQL
1:18:51
run actually spawns the possible SQL automatically so you can play around and
1:18:54
automatically so you can play around and
1:18:54
automatically so you can play around and then you create the extension and then
1:18:56
then you create the extension and then
1:18:56
then you create the extension and then you it has even has a built-in hello
1:18:58
you it has even has a built-in hello
1:18:58
you it has even has a built-in hello world
1:18:59
world
1:18:59
world so
1:19:00
so
1:19:00
so you have you have an extension that
1:19:02
you have you have an extension that
1:19:02
you have you have an extension that works it doesn't do very useful things
1:19:03
works it doesn't do very useful things
1:19:03
works it doesn't do very useful things yet but it works
1:19:07
so then I wanted to make some changes
1:19:08
so then I wanted to make some changes
1:19:08
so then I wanted to make some changes because I mean I wanted to make it do
1:19:10
because I mean I wanted to make it do
1:19:10
because I mean I wanted to make it do useful thing so I made some small
1:19:11
useful thing so I made some small
1:19:11
useful thing so I made some small changes I recompiled and then it took
1:19:13
changes I recompiled and then it took
1:19:13
changes I recompiled and then it took seven seconds to compile so that's
1:19:15
seven seconds to compile so that's
1:19:15
seven seconds to compile so that's that's not really I mean it's not super
1:19:18
that's not really I mean it's not super
1:19:18
that's not really I mean it's not super slow but it's it's slow enough that it
1:19:20
slow but it's it's slow enough that it
1:19:20
slow but it's it's slow enough that it annoys me uh so I looked into it a bit
1:19:23
annoys me uh so I looked into it a bit
1:19:23
annoys me uh so I looked into it a bit like why is it so slow it turns out that
1:19:26
like why is it so slow it turns out that
1:19:26
like why is it so slow it turns out that one thing that makes rust compilation
1:19:28
one thing that makes rust compilation
1:19:28
one thing that makes rust compilation and compilation in general slow usually
1:19:31
and compilation in general slow usually
1:19:31
and compilation in general slow usually is linking which is compilers they
1:19:34
is linking which is compilers they
1:19:34
is linking which is compilers they create multiple files and then they sort
1:19:37
create multiple files and then they sort
1:19:37
create multiple files and then they sort of merge them together in a final binary
1:19:39
of merge them together in a final binary
1:19:39
of merge them together in a final binary that you can
1:19:41
that you can
1:19:41
that you can and that step it's it's don't buy the
1:19:44
and that step it's it's don't buy the
1:19:44
and that step it's it's don't buy the link combining those files and that's
1:19:47
link combining those files and that's
1:19:47
link combining those files and that's and the new Linker which is sort of the
1:19:49
and the new Linker which is sort of the
1:19:49
and the new Linker which is sort of the default one that you get on Linux it's a
1:19:52
default one that you get on Linux it's a
1:19:52
default one that you get on Linux it's a bit slow
1:19:54
bit slow
1:19:54
bit slow so
1:19:55
so
1:19:55
so it's very easy you can swap it out for a
1:19:57
it's very easy you can swap it out for a
1:19:57
it's very easy you can swap it out for a faster one mode is fast ldd is fast uh
1:20:01
faster one mode is fast ldd is fast uh
1:20:01
faster one mode is fast ldd is fast uh LED is from the Clank people
1:20:03
LED is from the Clank people
1:20:03
LED is from the Clank people and multis from someone else but uh I
1:20:07
and multis from someone else but uh I
1:20:07
and multis from someone else but uh I tried molt and it was like ah so the
1:20:09
tried molt and it was like ah so the
1:20:09
tried molt and it was like ah so the three three lines in some file in my
1:20:11
three three lines in some file in my
1:20:11
three three lines in some file in my home directory and then
1:20:13
home directory and then
1:20:13
home directory and then and then it took only one second to
1:20:15
and then it took only one second to
1:20:15
and then it took only one second to compile instead of seven so that's okay
1:20:18
compile instead of seven so that's okay
1:20:18
compile instead of seven so that's okay one second then I think is acceptable
1:20:20
one second then I think is acceptable
1:20:20
one second then I think is acceptable that's that's that's fine my uh my
1:20:23
that's that's that's fine my uh my
1:20:23
that's that's that's fine my uh my change change compile cycle is it's not
1:20:25
change change compile cycle is it's not
1:20:25
change change compile cycle is it's not impacted so much by that
1:20:27
impacted so much by that
1:20:27
impacted so much by that so now we can actually start uh so what
1:20:30
so now we can actually start uh so what
1:20:30
so now we can actually start uh so what do we actually need in this extension we
1:20:32
do we actually need in this extension we
1:20:32
do we actually need in this extension we first we need to get the schema
1:20:34
first we need to get the schema
1:20:34
first we need to get the schema definition that was sort of the whole
1:20:35
definition that was sort of the whole
1:20:35
definition that was sort of the whole point we wanted to avoid copy pasting
1:20:37
point we wanted to avoid copy pasting
1:20:37
point we wanted to avoid copy pasting that schema definition and then we need
1:20:39
that schema definition and then we need
1:20:39
that schema definition and then we need to send it to gbt because that's sort of
1:20:41
to send it to gbt because that's sort of
1:20:41
to send it to gbt because that's sort of the face part
1:20:42
the face part
1:20:42
the face part and then I mean we got some query that
1:20:45
and then I mean we got some query that
1:20:45
and then I mean we got some query that that we can might execute and we kind of
1:20:49
that we can might execute and we kind of
1:20:49
that we can might execute and we kind of don't want to copy paste that either so
1:20:51
don't want to copy paste that either so
1:20:51
don't want to copy paste that either so we just want to execute that I mean I'm
1:20:53
we just want to execute that I mean I'm
1:20:53
we just want to execute that I mean I'm not going to read it anyway that's
1:20:56
not going to read it anyway that's
1:20:56
not going to read it anyway that's that's way too much work
1:20:58
that's way too much work
1:20:58
that's way too much work so I'll I'll just I'll just trust that
1:21:00
so I'll I'll just I'll just trust that
1:21:00
so I'll I'll just I'll just trust that it sort of does what I want and uh
1:21:03
it sort of does what I want and uh
1:21:03
it sort of does what I want and uh and execute it right away and then if
1:21:06
and execute it right away and then if
1:21:06
and execute it right away and then if that query is like a select where you
1:21:08
that query is like a select where you
1:21:08
that query is like a select where you want to
1:21:09
want to
1:21:09
want to factual data
1:21:11
factual data
1:21:11
factual data that that should actually also return
1:21:14
that that should actually also return
1:21:14
that that should actually also return that data not just run and then throw
1:21:17
that data not just run and then throw
1:21:17
that data not just run and then throw everything away so we need to sort of
1:21:18
everything away so we need to sort of
1:21:18
everything away so we need to sort of return arbitrary SQL results because all
1:21:21
return arbitrary SQL results because all
1:21:21
return arbitrary SQL results because all the queries they can
1:21:22
the queries they can
1:21:22
the queries they can I mean you don't know what kind of query
1:21:24
I mean you don't know what kind of query
1:21:24
I mean you don't know what kind of query it is it kind of many columns or just
1:21:27
it is it kind of many columns or just
1:21:27
it is it kind of many columns or just one uh that's yeah
1:21:30
one uh that's yeah
1:21:30
one uh that's yeah so let's start and getting the schema
1:21:33
so let's start and getting the schema
1:21:33
so let's start and getting the schema definition is actually not super easy
1:21:35
definition is actually not super easy
1:21:35
definition is actually not super easy postgres somehow doesn't have like a
1:21:37
postgres somehow doesn't have like a
1:21:37
postgres somehow doesn't have like a built-in way to do this which
1:21:39
built-in way to do this which
1:21:39
built-in way to do this which is very surprising to me so
1:21:42
is very surprising to me so
1:21:42
is very surprising to me so um
1:21:43
um
1:21:43
um well it's also not the worst there's
1:21:45
well it's also not the worst there's
1:21:45
well it's also not the worst there's like sort of internal tables from
1:21:47
like sort of internal tables from
1:21:47
like sort of internal tables from postgres that you can look at uh to see
1:21:50
postgres that you can look at uh to see
1:21:50
postgres that you can look at uh to see what tables exist
1:21:52
what tables exist
1:21:52
what tables exist and I just did that I looked at the
1:21:54
and I just did that I looked at the
1:21:54
and I just did that I looked at the column names it sort of concatenated
1:21:56
column names it sort of concatenated
1:21:56
column names it sort of concatenated strings together to
1:21:58
strings together to
1:21:58
strings together to turn it into something that looked like
1:22:00
turn it into something that looked like
1:22:00
turn it into something that looked like create table statements
1:22:03
create table statements
1:22:03
create table statements and then I had like a schema the schema
1:22:06
and then I had like a schema the schema
1:22:06
and then I had like a schema the schema description
1:22:07
description
1:22:07
description and then I need to send them to chatgpt
1:22:09
and then I need to send them to chatgpt
1:22:09
and then I need to send them to chatgpt and it's very easy to do that you can
1:22:11
and it's very easy to do that you can
1:22:11
and it's very easy to do that you can install the open AI package for rust and
1:22:15
install the open AI package for rust and
1:22:15
install the open AI package for rust and cargo is the package menu again so it's
1:22:17
cargo is the package menu again so it's
1:22:17
cargo is the package menu again so it's like a one one line command and you have
1:22:20
like a one one line command and you have
1:22:20
like a one one line command and you have you have a 1050 client that you can send
1:22:24
you have a 1050 client that you can send
1:22:24
you have a 1050 client that you can send questions to
1:22:26
questions to
1:22:26
questions to and then you need to make those
1:22:28
and then you need to make those
1:22:28
and then you need to make those questions so we first start with giving
1:22:30
questions so we first start with giving
1:22:30
questions so we first start with giving the AI a little bit of confidence you
1:22:32
the AI a little bit of confidence you
1:22:32
the AI a little bit of confidence you wanted to to really give you the good
1:22:34
wanted to to really give you the good
1:22:34
wanted to to really give you the good answers and it only does that if you if
1:22:35
answers and it only does that if you if
1:22:36
answers and it only does that if you if you tell it they're an expert so they're
1:22:38
you tell it they're an expert so they're
1:22:38
you tell it they're an expert so they're like a postgres expert otherwise I mean
1:22:40
like a postgres expert otherwise I mean
1:22:40
like a postgres expert otherwise I mean maybe they're just saying dumb things
1:22:41
maybe they're just saying dumb things
1:22:41
maybe they're just saying dumb things but if they're an expert then they will
1:22:43
but if they're an expert then they will
1:22:43
but if they're an expert then they will never do
1:22:44
never do
1:22:45
never do so that's one thing then you want to
1:22:47
so that's one thing then you want to
1:22:47
so that's one thing then you want to tell it what you have like this is my
1:22:49
tell it what you have like this is my
1:22:49
tell it what you have like this is my database schema so that's the second
1:22:51
database schema so that's the second
1:22:51
database schema so that's the second thing you sent it to
1:22:53
thing you sent it to
1:22:53
thing you sent it to and then you tell it what you need like
1:22:55
and then you tell it what you need like
1:22:55
and then you tell it what you need like after schema
1:22:57
after schema
1:22:57
after schema and I want you to give me a query that
1:22:59
and I want you to give me a query that
1:22:59
and I want you to give me a query that that does this normal human description
1:23:03
that does this normal human description
1:23:03
that does this normal human description thing
1:23:05
and then finally
1:23:06
and then finally
1:23:06
and then finally also really important you tell it what
1:23:08
also really important you tell it what
1:23:08
also really important you tell it what you don't want so you don't want to kind
1:23:10
you don't want so you don't want to kind
1:23:10
you don't want so you don't want to kind of fluff around it with with things that
1:23:13
of fluff around it with with things that
1:23:13
of fluff around it with with things that that you don't care about like your sort
1:23:16
that you don't care about like your sort
1:23:16
that you don't care about like your sort of descriptions I mean you just want the
1:23:17
of descriptions I mean you just want the
1:23:17
of descriptions I mean you just want the raw SQL query that's that's all you need
1:23:19
raw SQL query that's that's all you need
1:23:19
raw SQL query that's that's all you need and you don't want it to make up tables
1:23:21
and you don't want it to make up tables
1:23:21
and you don't want it to make up tables that don't exist I mean the things
1:23:22
that don't exist I mean the things
1:23:22
that don't exist I mean the things you're sending it it's all the tables
1:23:24
you're sending it it's all the tables
1:23:24
you're sending it it's all the tables that are there so if it if it thinks
1:23:26
that are there so if it if it thinks
1:23:26
that are there so if it if it thinks there's other tables it's like okay well
1:23:27
there's other tables it's like okay well
1:23:27
there's other tables it's like okay well that I mean that query is never going to
1:23:29
that I mean that query is never going to
1:23:29
that I mean that query is never going to work anyway
1:23:30
work anyway
1:23:30
work anyway so you try to
1:23:32
so you try to
1:23:32
so you try to suggest to it that it shouldn't do that
1:23:34
suggest to it that it shouldn't do that
1:23:34
suggest to it that it shouldn't do that and that's that's the important part it
1:23:36
and that's that's the important part it
1:23:36
and that's that's the important part it never really completely listens so most
1:23:39
never really completely listens so most
1:23:39
never really completely listens so most of the time it does but sometimes
1:23:40
of the time it does but sometimes
1:23:40
of the time it does but sometimes sometimes it's just it's just ignores
1:23:42
sometimes it's just it's just ignores
1:23:42
sometimes it's just it's just ignores your advice
1:23:44
your advice
1:23:44
your advice but yeah that's I mean people also do
1:23:48
but yeah that's I mean people also do
1:23:48
but yeah that's I mean people also do but there's one issue with sending these
1:23:50
but there's one issue with sending these
1:23:50
but there's one issue with sending these requests and that's that
1:23:54
rushed it loves async and threads to
1:23:57
rushed it loves async and threads to
1:23:57
rushed it loves async and threads to make everything go fast
1:23:59
make everything go fast
1:23:59
make everything go fast and async support in PGX it doesn't
1:24:01
and async support in PGX it doesn't
1:24:01
and async support in PGX it doesn't really exist
1:24:03
really exist
1:24:03
really exist I opened an issue and get up to sort of
1:24:05
I opened an issue and get up to sort of
1:24:05
I opened an issue and get up to sort of discuss a bit how we could make that
1:24:06
discuss a bit how we could make that
1:24:06
discuss a bit how we could make that better but at the moment it doesn't it
1:24:08
better but at the moment it doesn't it
1:24:08
better but at the moment it doesn't it there's no there's not really a nice
1:24:09
there's no there's not really a nice
1:24:09
there's no there's not really a nice solution and postgres and PGX really
1:24:12
solution and postgres and PGX really
1:24:12
solution and postgres and PGX really don't like threads nothing in postgres
1:24:15
don't like threads nothing in postgres
1:24:15
don't like threads nothing in postgres is stretch safe so if you spawn multiple
1:24:18
is stretch safe so if you spawn multiple
1:24:18
is stretch safe so if you spawn multiple threats with touch postgres things it's
1:24:20
threats with touch postgres things it's
1:24:20
threats with touch postgres things it's gonna everything is gonna go horribly
1:24:22
gonna everything is gonna go horribly
1:24:22
gonna everything is gonna go horribly wrong
1:24:23
wrong
1:24:23
wrong but luckily there's a really easy to
1:24:26
but luckily there's a really easy to
1:24:26
but luckily there's a really easy to work around you can just tell like the
1:24:28
work around you can just tell like the
1:24:28
work around you can just tell like the the Rest async Library to run everything
1:24:31
the Rest async Library to run everything
1:24:31
the Rest async Library to run everything on the current thread and then you don't
1:24:33
on the current thread and then you don't
1:24:33
on the current thread and then you don't use threats it won't go as fast as it
1:24:35
use threats it won't go as fast as it
1:24:35
use threats it won't go as fast as it could but that's fine we're only sending
1:24:37
could but that's fine we're only sending
1:24:37
could but that's fine we're only sending one request at the time anyway
1:24:38
one request at the time anyway
1:24:38
one request at the time anyway so we might as well do it on the current
1:24:40
so we might as well do it on the current
1:24:40
so we might as well do it on the current thread
1:24:43
and then we get some query back
1:24:46
and then we get some query back
1:24:46
and then we get some query back that we can execute and then we need to
1:24:48
that we can execute and then we need to
1:24:48
that we can execute and then we need to execute that and postgres SPI which is
1:24:50
execute that and postgres SPI which is
1:24:50
execute that and postgres SPI which is server programming interface which is
1:24:52
server programming interface which is
1:24:52
server programming interface which is like a c bindings for running SQL in an
1:24:56
like a c bindings for running SQL in an
1:24:56
like a c bindings for running SQL in an extension
1:24:58
extension
1:24:58
extension and PGX has really nice wrist bindings
1:25:00
and PGX has really nice wrist bindings
1:25:00
and PGX has really nice wrist bindings for those C bindings so it's just as
1:25:03
for those C bindings so it's just as
1:25:03
for those C bindings so it's just as simple as you get a string of SQL and
1:25:05
simple as you get a string of SQL and
1:25:05
simple as you get a string of SQL and you say execute connect SBI and then
1:25:09
you say execute connect SBI and then
1:25:09
you say execute connect SBI and then executes this uh this sequel that's all
1:25:12
executes this uh this sequel that's all
1:25:12
executes this uh this sequel that's all you need to do
1:25:13
you need to do
1:25:13
you need to do so that's this was actually very simple
1:25:15
so that's this was actually very simple
1:25:15
so that's this was actually very simple but the returning arbitrary results is
1:25:17
but the returning arbitrary results is
1:25:17
but the returning arbitrary results is where it gets a bit more complicated
1:25:19
where it gets a bit more complicated
1:25:19
where it gets a bit more complicated because
1:25:21
because
1:25:21
because all you can do with the postgres
1:25:22
all you can do with the postgres
1:25:22
all you can do with the postgres extension is create functions and
1:25:24
extension is create functions and
1:25:24
extension is create functions and functions they need to have a known set
1:25:27
functions they need to have a known set
1:25:27
functions they need to have a known set of columns you can't change those
1:25:29
of columns you can't change those
1:25:29
of columns you can't change those columns on the fly or the while running
1:25:31
columns on the fly or the while running
1:25:31
columns on the fly or the while running the function postgres needs to know what
1:25:34
the function postgres needs to know what
1:25:34
the function postgres needs to know what what columns that function is going to
1:25:36
what columns that function is going to
1:25:36
what columns that function is going to return return
1:25:37
return return
1:25:37
return return so
1:25:39
so
1:25:39
so that's that seems a bit of a problem but
1:25:41
that's that seems a bit of a problem but
1:25:41
that's that seems a bit of a problem but actually it's not so much
1:25:44
actually it's not so much
1:25:44
actually it's not so much we are not running any SQL already
1:25:46
we are not running any SQL already
1:25:46
we are not running any SQL already anyway so we're sort of doing nosql kind
1:25:48
anyway so we're sort of doing nosql kind
1:25:48
anyway so we're sort of doing nosql kind of things so we might as well go through
1:25:49
of things so we might as well go through
1:25:49
of things so we might as well go through you know SQL and start returning Json
1:25:51
you know SQL and start returning Json
1:25:51
you know SQL and start returning Json blobs so that's what we're doing we're
1:25:54
blobs so that's what we're doing we're
1:25:54
blobs so that's what we're doing we're we're postgres can
1:25:57
we're postgres can
1:25:57
we're postgres can transform a query and return one Json
1:25:59
transform a query and return one Json
1:25:59
transform a query and return one Json blob so we have a function that returns
1:26:01
blob so we have a function that returns
1:26:01
blob so we have a function that returns only Json like a single column
1:26:04
only Json like a single column
1:26:04
only Json like a single column with only Json
1:26:06
with only Json
1:26:06
with only Json and that that's kind of I mean and you
1:26:10
and that that's kind of I mean and you
1:26:10
and that that's kind of I mean and you and you so and you wrap the actual query
1:26:12
and you so and you wrap the actual query
1:26:12
and you so and you wrap the actual query that you get from jet PPT in like a
1:26:13
that you get from jet PPT in like a
1:26:13
that you get from jet PPT in like a little bit of wrapper to turn that query
1:26:15
little bit of wrapper to turn that query
1:26:15
little bit of wrapper to turn that query into returning Json
1:26:17
into returning Json
1:26:17
into returning Json and then then you have something that
1:26:19
and then then you have something that
1:26:19
and then then you have something that works
1:26:20
works
1:26:20
works so now it's time to show that all this
1:26:24
so now it's time to show that all this
1:26:24
so now it's time to show that all this sort of does what they tell it it does
1:26:26
sort of does what they tell it it does
1:26:26
sort of does what they tell it it does so we're gonna hope that AI is uh doing
1:26:30
so we're gonna hope that AI is uh doing
1:26:30
so we're gonna hope that AI is uh doing the right thing but I didn't know really
1:26:32
the right thing but I didn't know really
1:26:32
the right thing but I didn't know really what kind of demo I needed to do
1:26:34
what kind of demo I needed to do
1:26:34
what kind of demo I needed to do so I'll ask jtp again I was like ah to
1:26:37
so I'll ask jtp again I was like ah to
1:26:37
so I'll ask jtp again I was like ah to do list that that's a good demo app I
1:26:39
do list that that's a good demo app I
1:26:39
do list that that's a good demo app I was like okay well that seems that seems
1:26:41
was like okay well that seems that seems
1:26:41
was like okay well that seems that seems reasonable so let's uh let's do that I
1:26:45
reasonable so let's uh let's do that I
1:26:45
reasonable so let's uh let's do that I will uh
1:26:47
will uh
1:26:47
will uh I will start typing now and uh so the
1:26:51
I will start typing now and uh so the
1:26:51
I will start typing now and uh so the thing I do is let's first create some
1:26:54
thing I do is let's first create some
1:26:54
thing I do is let's first create some tables and as you can see
1:26:56
tables and as you can see
1:26:56
tables and as you can see um I'm calling this function I'm very
1:26:58
um I'm calling this function I'm very
1:26:58
um I'm calling this function I'm very I'm feeling very lucky so it's a I mean
1:27:01
I'm feeling very lucky so it's a I mean
1:27:01
I'm feeling very lucky so it's a I mean this is changing your database with
1:27:03
this is changing your database with
1:27:03
this is changing your database with something returned from an AI that might
1:27:06
something returned from an AI that might
1:27:06
something returned from an AI that might or might not do exactly what you want so
1:27:08
or might not do exactly what you want so
1:27:08
or might not do exactly what you want so I think it's a well-named function
1:27:10
I think it's a well-named function
1:27:10
I think it's a well-named function and what we're going to do is we're
1:27:11
and what we're going to do is we're
1:27:12
and what we're going to do is we're creating tables for to-do app with
1:27:14
creating tables for to-do app with
1:27:14
creating tables for to-do app with multiple users
1:27:16
multiple users
1:27:16
multiple users and
1:27:17
and
1:27:17
and when I run that
1:27:21
it's uh it's going to think for a little
1:27:23
it's uh it's going to think for a little
1:27:23
it's uh it's going to think for a little bit
1:27:24
bit
1:27:24
bit and if we scroll up a bit you can see it
1:27:26
and if we scroll up a bit you can see it
1:27:26
and if we scroll up a bit you can see it created a table with users
1:27:28
created a table with users
1:27:28
created a table with users and it created table with tasks
1:27:30
and it created table with tasks
1:27:30
and it created table with tasks and user ID description due date is
1:27:34
and user ID description due date is
1:27:34
and user ID description due date is complete for all the tasks and names and
1:27:36
complete for all the tasks and names and
1:27:36
complete for all the tasks and names and emails so that looks that looks pretty
1:27:37
emails so that looks that looks pretty
1:27:37
emails so that looks that looks pretty reasonable but the due date seems a bit
1:27:40
reasonable but the due date seems a bit
1:27:40
reasonable but the due date seems a bit unnecessary
1:27:42
unnecessary
1:27:42
unnecessary so let's let's remove that column we
1:27:45
so let's let's remove that column we
1:27:45
so let's let's remove that column we don't really care about it uh enter GPT
1:27:48
don't really care about it uh enter GPT
1:27:48
don't really care about it uh enter GPT happily obliges it's uh it runs another
1:27:51
happily obliges it's uh it runs another
1:27:51
happily obliges it's uh it runs another table drop column due date even though
1:27:53
table drop column due date even though
1:27:53
table drop column due date even though we didn't specify what table it was
1:27:57
we didn't specify what table it was
1:27:57
we didn't specify what table it was um
1:27:58
um
1:27:58
um or that there was an underscore between
1:28:00
or that there was an underscore between
1:28:00
or that there was an underscore between Jew and date it sort of
1:28:02
Jew and date it sort of
1:28:02
Jew and date it sort of it sort of understands what we mean
1:28:05
it sort of understands what we mean
1:28:05
it sort of understands what we mean so and if I look at the tables that we
1:28:08
so and if I look at the tables that we
1:28:08
so and if I look at the tables that we have we did have the task in the user
1:28:09
have we did have the task in the user
1:28:09
have we did have the task in the user stable
1:28:10
stable
1:28:10
stable so now we need some data data to show
1:28:13
so now we need some data data to show
1:28:13
so now we need some data data to show some nice queries and creating dummy
1:28:16
some nice queries and creating dummy
1:28:16
some nice queries and creating dummy data is always quite annoying that you
1:28:19
data is always quite annoying that you
1:28:19
data is always quite annoying that you have to think and I mean certificate can
1:28:22
have to think and I mean certificate can
1:28:22
have to think and I mean certificate can take for us so let's
1:28:26
let's create
1:28:30
um let's create
1:28:31
um let's create
1:28:31
um let's create let's add three famous football players
1:28:33
let's add three famous football players
1:28:33
let's add three famous football players and have three tasks for each of them
1:28:35
and have three tasks for each of them
1:28:35
and have three tasks for each of them based on the training schedule to have
1:28:37
based on the training schedule to have
1:28:37
based on the training schedule to have some nice relatable data
1:28:40
some nice relatable data
1:28:40
some nice relatable data and we wait a bit this usually takes a
1:28:42
and we wait a bit this usually takes a
1:28:42
and we wait a bit this usually takes a bit longer because it actually
1:28:43
bit longer because it actually
1:28:43
bit longer because it actually it's it's
1:28:45
it's it's
1:28:45
it's it's creates
1:28:47
creates
1:28:47
creates tokens and those each of those Dockers
1:28:49
tokens and those each of those Dockers
1:28:49
tokens and those each of those Dockers cost a bit of time but this time it was
1:28:51
cost a bit of time but this time it was
1:28:51
cost a bit of time but this time it was fairly fast only seven seconds uh but
1:28:53
fairly fast only seven seconds uh but
1:28:53
fairly fast only seven seconds uh but because this is I mean this it's
1:28:55
because this is I mean this it's
1:28:55
because this is I mean this it's actually creating quite a bit of data so
1:28:57
actually creating quite a bit of data so
1:28:57
actually creating quite a bit of data so it takes more time to generate that than
1:28:59
it takes more time to generate that than
1:28:59
it takes more time to generate that than a simple query
1:29:01
a simple query
1:29:01
a simple query and as you can see it's like it adds uh
1:29:03
and as you can see it's like it adds uh
1:29:03
and as you can see it's like it adds uh Messi and Ronaldo and Neymar I mean
1:29:05
Messi and Ronaldo and Neymar I mean
1:29:06
Messi and Ronaldo and Neymar I mean another huge football fan or anything
1:29:07
another huge football fan or anything
1:29:07
another huge football fan or anything but I mean those even even for me those
1:29:09
but I mean those even even for me those
1:29:09
but I mean those even even for me those are names that I know and it even knows
1:29:12
are names that I know and it even knows
1:29:12
are names that I know and it even knows sort of the football clubs that they're
1:29:14
sort of the football clubs that they're
1:29:14
sort of the football clubs that they're from and uses those in the email so
1:29:16
from and uses those in the email so
1:29:16
from and uses those in the email so that's it's really it's really nice
1:29:17
that's it's really it's really nice
1:29:17
that's it's really it's really nice touch
1:29:18
touch
1:29:18
touch and and uh the past they also sort of
1:29:21
and and uh the past they also sort of
1:29:21
and and uh the past they also sort of make make sense
1:29:23
make make sense
1:29:23
make make sense um they they look like football things
1:29:25
um they they look like football things
1:29:25
um they they look like football things that you could train
1:29:28
so yeah uh let's see if any anyone
1:29:31
so yeah uh let's see if any anyone
1:29:31
so yeah uh let's see if any anyone completes some of their tasks
1:29:36
but if there's any completed tasks you
1:29:38
but if there's any completed tasks you
1:29:38
but if there's any completed tasks you can you can check like the public and
1:29:41
can you can check like the public and
1:29:41
can you can check like the public and and
1:29:42
and
1:29:42
and it's uh it will create a query it does
1:29:44
it's uh it will create a query it does
1:29:45
it's uh it will create a query it does is complete it's true
1:29:47
is complete it's true
1:29:47
is complete it's true and but there's currently no one that
1:29:48
and but there's currently no one that
1:29:48
and but there's currently no one that completed any tasks so let's let's
1:29:50
completed any tasks so let's let's
1:29:50
completed any tasks so let's let's change that let's add some
1:29:53
change that let's add some
1:29:53
change that let's add some random
1:29:54
random
1:29:54
random let's complete some tasks randomly so we
1:29:57
let's complete some tasks randomly so we
1:29:57
let's complete some tasks randomly so we randomly complete 50 of the tasks
1:30:01
randomly complete 50 of the tasks
1:30:01
randomly complete 50 of the tasks and it takes a bit of time but uh then
1:30:05
and it takes a bit of time but uh then
1:30:05
and it takes a bit of time but uh then we got a query back like updates tasks
1:30:08
we got a query back like updates tasks
1:30:08
we got a query back like updates tasks set is complete to true and then where I
1:30:11
set is complete to true and then where I
1:30:11
set is complete to true and then where I be in uh something I don't know so some
1:30:14
be in uh something I don't know so some
1:30:14
be in uh something I don't know so some query uh with some random parts so that
1:30:17
query uh with some random parts so that
1:30:17
query uh with some random parts so that seems reasonable time 0.5 so that's
1:30:19
seems reasonable time 0.5 so that's
1:30:19
seems reasonable time 0.5 so that's probably the sort of the chance of it
1:30:21
probably the sort of the chance of it
1:30:21
probably the sort of the chance of it happening so that I mean I I don't know
1:30:23
happening so that I mean I I don't know
1:30:24
happening so that I mean I I don't know the exact details but it looks looks
1:30:25
the exact details but it looks looks
1:30:25
the exact details but it looks looks about all right
1:30:27
about all right
1:30:27
about all right um so let's let's take a look again
1:30:29
um so let's let's take a look again
1:30:30
um so let's let's take a look again at uh
1:30:32
at uh
1:30:32
at uh let's show the completed tasks again oh
1:30:34
let's show the completed tasks again oh
1:30:34
let's show the completed tasks again oh yeah you can see like oh there's five
1:30:36
yeah you can see like oh there's five
1:30:36
yeah you can see like oh there's five completed tasks now they don't they
1:30:38
completed tasks now they don't they
1:30:38
completed tasks now they don't they might not be entirely random because
1:30:39
might not be entirely random because
1:30:39
might not be entirely random because it's like five six seven eight nine but
1:30:41
it's like five six seven eight nine but
1:30:41
it's like five six seven eight nine but maybe you know it could be could be
1:30:43
maybe you know it could be could be
1:30:43
maybe you know it could be could be random
1:30:45
random
1:30:45
random and now but now we kind of also want to
1:30:47
and now but now we kind of also want to
1:30:47
and now but now we kind of also want to know
1:30:48
know
1:30:48
know if
1:30:50
if
1:30:50
if if they if who's who's complete these
1:30:53
if they if who's who's complete these
1:30:53
if they if who's who's complete these steps and we only see user ID two and
1:30:55
steps and we only see user ID two and
1:30:55
steps and we only see user ID two and three so that's a bit annoying uh so
1:30:58
three so that's a bit annoying uh so
1:30:58
three so that's a bit annoying uh so let's let's use Let's do let's show the
1:31:01
let's let's use Let's do let's show the
1:31:01
let's let's use Let's do let's show the amount of completed tasks by user and
1:31:03
amount of completed tasks by user and
1:31:03
amount of completed tasks by user and include the username just so that
1:31:05
include the username just so that
1:31:05
include the username just so that we know what what what's uh football
1:31:08
we know what what what's uh football
1:31:08
we know what what what's uh football players actually doing their trading and
1:31:10
players actually doing their trading and
1:31:10
players actually doing their trading and which one is not doing anything
1:31:12
which one is not doing anything
1:31:12
which one is not doing anything so we wait a bit again and we can see
1:31:15
so we wait a bit again and we can see
1:31:15
so we wait a bit again and we can see that Neymar is clearly the one that's
1:31:17
that Neymar is clearly the one that's
1:31:17
that Neymar is clearly the one that's that's actually training a lot and
1:31:19
that's actually training a lot and
1:31:19
that's actually training a lot and Ronaldo's behind but Messi is is just I
1:31:21
Ronaldo's behind but Messi is is just I
1:31:21
Ronaldo's behind but Messi is is just I mean he's he's probably old and not I
1:31:23
mean he's he's probably old and not I
1:31:23
mean he's he's probably old and not I mean he thinks he's done enough
1:31:27
so we have
1:31:28
so we have
1:31:28
so we have and and the query it's actually it's
1:31:30
and and the query it's actually it's
1:31:30
and and the query it's actually it's actually not a super easy query it has
1:31:32
actually not a super easy query it has
1:31:32
actually not a super easy query it has like a join on like the columns that are
1:31:35
like a join on like the columns that are
1:31:35
like a join on like the columns that are actually there and it does a group by
1:31:37
actually there and it does a group by
1:31:37
actually there and it does a group by and it thinks of all these I mean it
1:31:39
and it thinks of all these I mean it
1:31:39
and it thinks of all these I mean it does all the things sort of
1:31:41
does all the things sort of
1:31:41
does all the things sort of in a way that's that makes sense so
1:31:44
in a way that's that makes sense so
1:31:44
in a way that's that makes sense so that's it's quite impressive to me
1:31:47
that's it's quite impressive to me
1:31:47
that's it's quite impressive to me and that's kind of the demo but I one
1:31:50
and that's kind of the demo but I one
1:31:50
and that's kind of the demo but I one final thing is usually after them I
1:31:52
final thing is usually after them I
1:31:52
final thing is usually after them I usually like to clean up uh because I
1:31:55
usually like to clean up uh because I
1:31:55
usually like to clean up uh because I don't like leaving tables around so they
1:31:56
don't like leaving tables around so they
1:31:56
don't like leaving tables around so they mess up other tests that I do
1:31:59
mess up other tests that I do
1:31:59
mess up other tests that I do so let's let's uh sort of uh remove uh
1:32:02
so let's let's uh sort of uh remove uh
1:32:02
so let's let's uh sort of uh remove uh with the leads let's just delete all the
1:32:05
with the leads let's just delete all the
1:32:05
with the leads let's just delete all the tables
1:32:06
tables
1:32:06
tables uh at the end so we drop tables and it
1:32:09
uh at the end so we drop tables and it
1:32:09
uh at the end so we drop tables and it deletes the tables very happily so this
1:32:11
deletes the tables very happily so this
1:32:11
deletes the tables very happily so this is one of the reasons why you probably
1:32:12
is one of the reasons why you probably
1:32:12
is one of the reasons why you probably shouldn't do this all the time uh on
1:32:16
shouldn't do this all the time uh on
1:32:16
shouldn't do this all the time uh on your production database uh I mean
1:32:18
your production database uh I mean
1:32:18
your production database uh I mean whatever you tell chat update uh it will
1:32:21
whatever you tell chat update uh it will
1:32:21
whatever you tell chat update uh it will execute and it's totally totally fine
1:32:25
execute and it's totally totally fine
1:32:25
execute and it's totally totally fine uh that's that's uh that was a demo
1:32:33
and so where can you find this amazing
1:32:35
and so where can you find this amazing
1:32:35
and so where can you find this amazing postgres extension it's uh open source I
1:32:38
postgres extension it's uh open source I
1:32:38
postgres extension it's uh open source I just published it like five minutes
1:32:40
just published it like five minutes
1:32:40
just published it like five minutes before my talk
1:32:41
before my talk
1:32:41
before my talk uh definitely use it your own risk it's
1:32:43
uh definitely use it your own risk it's
1:32:43
uh definitely use it your own risk it's it's sort of it's more toy than
1:32:45
it's sort of it's more toy than
1:32:45
it's sort of it's more toy than production project at this uh at this
1:32:48
production project at this uh at this
1:32:48
production project at this uh at this moment
1:32:49
moment
1:32:49
moment uh but it's it's fun to look at the code
1:32:51
uh but it's it's fun to look at the code
1:32:51
uh but it's it's fun to look at the code is quite quite easily understandable I
1:32:53
is quite quite easily understandable I
1:32:53
is quite quite easily understandable I think
1:32:54
think
1:32:54
think that's uh yeah and there's a few things
1:32:57
that's uh yeah and there's a few things
1:32:57
that's uh yeah and there's a few things that for future improvements or like
1:32:59
that for future improvements or like
1:32:59
that for future improvements or like future fun things to maybe add to it
1:33:01
future fun things to maybe add to it
1:33:01
future fun things to maybe add to it like explaining explain plans because
1:33:03
like explaining explain plans because
1:33:03
like explaining explain plans because those are usually hard to understand
1:33:04
those are usually hard to understand
1:33:04
those are usually hard to understand index suggestions like what columns to
1:33:07
index suggestions like what columns to
1:33:07
index suggestions like what columns to index for your queries
1:33:08
index for your queries
1:33:08
index for your queries uh and also for psych this uh it's like
1:33:12
uh and also for psych this uh it's like
1:33:12
uh and also for psych this uh it's like distributed columns distribution column
1:33:14
distributed columns distribution column
1:33:14
distributed columns distribution column suggestions is another thing it's sort
1:33:16
suggestions is another thing it's sort
1:33:16
suggestions is another thing it's sort of even harder than index suggestions
1:33:17
of even harder than index suggestions
1:33:17
of even harder than index suggestions because you can only have one
1:33:18
because you can only have one
1:33:18
because you can only have one distribution column
1:33:20
distribution column
1:33:20
distribution column and it's it's not always the primary key
1:33:22
and it's it's not always the primary key
1:33:22
and it's it's not always the primary key it's uh you have you have to do some you
1:33:25
it's uh you have you have to do some you
1:33:25
it's uh you have you have to do some you have to think a bit about your skin line
1:33:27
have to think a bit about your skin line
1:33:27
have to think a bit about your skin line your and your queries to do something uh
1:33:30
your and your queries to do something uh
1:33:30
your and your queries to do something uh intelligent
1:33:32
intelligent
1:33:33
intelligent that's uh
1:33:35
that's uh
1:33:35
that's uh that that was uh my my talk are there
1:33:38
that that was uh my my talk are there
1:33:38
that that was uh my my talk are there any uh if there's any questions I'm
1:33:40
any uh if there's any questions I'm
1:33:40
any uh if there's any questions I'm happy to answer some
1:33:43
happy to answer some
1:33:43
happy to answer some okay so that was a delight yeah oh my
1:33:46
okay so that was a delight yeah oh my
1:33:46
okay so that was a delight yeah oh my goodness okay so we want to know did you
1:33:48
goodness okay so we want to know did you
1:33:48
goodness okay so we want to know did you actually write the code or did you make
1:33:50
actually write the code or did you make
1:33:50
actually write the code or did you make chat GPT write it all for the extension
1:33:53
chat GPT write it all for the extension
1:33:53
chat GPT write it all for the extension so I I I I wrote it all there were some
1:33:56
so I I I I wrote it all there were some
1:33:56
so I I I I wrote it all there were some examples I I know I know one of my
1:33:58
examples I I know I know one of my
1:33:58
examples I I know I know one of my colleagues Marco he uh he did he did use
1:34:01
colleagues Marco he uh he did he did use
1:34:01
colleagues Marco he uh he did he did use charity to you to use PGA like to tell
1:34:04
charity to you to use PGA like to tell
1:34:04
charity to you to use PGA like to tell you to create some other extension sort
1:34:07
you to create some other extension sort
1:34:07
you to create some other extension sort of template uh so it's definitely it
1:34:09
of template uh so it's definitely it
1:34:09
of template uh so it's definitely it sort of knows about it it's used so
1:34:11
sort of knows about it it's used so
1:34:11
sort of knows about it it's used so that's uh that's uh yeah
1:34:13
that's uh that's uh yeah
1:34:13
that's uh that's uh yeah and you just mentioned Marco slaw who's
1:34:15
and you just mentioned Marco slaw who's
1:34:15
and you just mentioned Marco slaw who's giving the keynote talk in the emea live
1:34:17
giving the keynote talk in the emea live
1:34:17
giving the keynote talk in the emea live stream that's tomorrow Wednesday just
1:34:19
stream that's tomorrow Wednesday just
1:34:19
stream that's tomorrow Wednesday just throwing in that shout out there
1:34:21
throwing in that shout out there
1:34:21
throwing in that shout out there um all right Rob yeah I just I I loved
1:34:25
um all right Rob yeah I just I I loved
1:34:25
um all right Rob yeah I just I I loved at the end where you're like you know be
1:34:26
at the end where you're like you know be
1:34:26
at the end where you're like you know be careful what you tell gbt to do earlier
1:34:28
careful what you tell gbt to do earlier
1:34:28
careful what you tell gbt to do earlier when you you said like let's remove the
1:34:31
when you you said like let's remove the
1:34:31
when you you said like let's remove the the you know the column and it like
1:34:33
the you know the column and it like
1:34:33
the you know the column and it like drops the call and I thought like wait
1:34:34
drops the call and I thought like wait
1:34:34
drops the call and I thought like wait was he removing that from his select
1:34:36
was he removing that from his select
1:34:36
was he removing that from his select statement or did he mean from the table
1:34:38
statement or did he mean from the table
1:34:38
statement or did he mean from the table so I'm just waiting for the day when
1:34:39
so I'm just waiting for the day when
1:34:39
so I'm just waiting for the day when it's our first you know like delete
1:34:41
it's our first you know like delete
1:34:41
it's our first you know like delete without a where Clause because of chat
1:34:43
without a where Clause because of chat
1:34:43
without a where Clause because of chat GPT like that will eventually be a thing
1:34:45
GPT like that will eventually be a thing
1:34:45
GPT like that will eventually be a thing that somebody has to tell their boss I'm
1:34:47
that somebody has to tell their boss I'm
1:34:47
that somebody has to tell their boss I'm sure so yes yeah for sure that's uh
1:34:50
sure so yes yeah for sure that's uh
1:34:50
sure so yes yeah for sure that's uh that's going to happen on a slightly
1:34:52
that's going to happen on a slightly
1:34:52
that's going to happen on a slightly more serious note so now that you've
1:34:54
more serious note so now that you've
1:34:54
more serious note so now that you've kind of played around with this rust
1:34:55
kind of played around with this rust
1:34:55
kind of played around with this rust extension uh and by the way thing I
1:34:58
extension uh and by the way thing I
1:34:58
extension uh and by the way thing I learned that they just changed their
1:34:59
learned that they just changed their
1:34:59
learned that they just changed their name like earlier you know within the
1:35:01
name like earlier you know within the
1:35:01
name like earlier you know within the last 24 hours so I think that's good
1:35:02
last 24 hours so I think that's good
1:35:03
last 24 hours so I think that's good long term but uh interesting timing uh
1:35:06
long term but uh interesting timing uh
1:35:06
long term but uh interesting timing uh I'm curious when you would maybe
1:35:07
I'm curious when you would maybe
1:35:07
I'm curious when you would maybe recommend to folks to use rust uh for an
1:35:11
recommend to folks to use rust uh for an
1:35:11
recommend to folks to use rust uh for an extension versus trying to do it in C
1:35:14
extension versus trying to do it in C
1:35:14
extension versus trying to do it in C so I actually think it's it's
1:35:17
so I actually think it's it's
1:35:17
so I actually think it's it's easier in Rust sort of for everyone
1:35:19
easier in Rust sort of for everyone
1:35:19
easier in Rust sort of for everyone honestly I I mean unless you're really
1:35:22
honestly I I mean unless you're really
1:35:22
honestly I I mean unless you're really uh like a postgres core developer like
1:35:25
uh like a postgres core developer like
1:35:25
uh like a postgres core developer like that works on podcast a lot I think it's
1:35:28
that works on podcast a lot I think it's
1:35:28
that works on podcast a lot I think it's it's it's easier to to learn Russ and
1:35:31
it's it's easier to to learn Russ and
1:35:31
it's it's easier to to learn Russ and and even the setup it's the setup is so
1:35:33
and even the setup it's the setup is so
1:35:33
and even the setup it's the setup is so much simpler than having like a creating
1:35:35
much simpler than having like a creating
1:35:35
much simpler than having like a creating extension in C is you have to do lots of
1:35:38
extension in C is you have to do lots of
1:35:38
extension in C is you have to do lots of things you have to look at make files
1:35:39
things you have to look at make files
1:35:39
things you have to look at make files it's it's all very annoying and uh
1:35:43
it's it's all very annoying and uh
1:35:43
it's it's all very annoying and uh and that's hard if you don't if you
1:35:45
and that's hard if you don't if you
1:35:45
and that's hard if you don't if you don't have sort of experience with it
1:35:46
don't have sort of experience with it
1:35:46
don't have sort of experience with it and with rust it was a few commands so
1:35:48
and with rust it was a few commands so
1:35:48
and with rust it was a few commands so you had something set up and there's
1:35:49
you had something set up and there's
1:35:49
you had something set up and there's some examples and that you can look at
1:35:52
some examples and that you can look at
1:35:52
some examples and that you can look at in the in the repo so it's quite it's
1:35:55
in the in the repo so it's quite it's
1:35:55
in the in the repo so it's quite it's really I think it was easier to write an
1:35:57
really I think it was easier to write an
1:35:57
really I think it was easier to write an extension in roster it wasn't C
1:35:59
extension in roster it wasn't C
1:35:59
extension in roster it wasn't C so I know you work on the situs
1:36:01
so I know you work on the situs
1:36:01
so I know you work on the situs extension to postgres are you saying
1:36:03
extension to postgres are you saying
1:36:03
extension to postgres are you saying that if you were to start the slightest
1:36:06
that if you were to start the slightest
1:36:06
that if you were to start the slightest project today you would be proposing to
1:36:08
project today you would be proposing to
1:36:08
project today you would be proposing to implement it in rest
1:36:09
implement it in rest
1:36:09
implement it in rest I think the slide this side this one
1:36:12
I think the slide this side this one
1:36:12
I think the slide this side this one might be sort of a special case because
1:36:13
might be sort of a special case because
1:36:13
might be sort of a special case because it interacts so much with all the all
1:36:16
it interacts so much with all the all
1:36:16
it interacts so much with all the all the postgres things so it's kind of
1:36:17
the postgres things so it's kind of
1:36:17
the postgres things so it's kind of useful to to be able to copy paste on
1:36:19
useful to to be able to copy paste on
1:36:19
useful to to be able to copy paste on code sometimes uh that that is not
1:36:22
code sometimes uh that that is not
1:36:22
code sometimes uh that that is not really exposed in postgres but we kind
1:36:24
really exposed in postgres but we kind
1:36:24
really exposed in postgres but we kind of still need
1:36:26
of still need
1:36:26
of still need um so that and if you have rusted I mean
1:36:28
um so that and if you have rusted I mean
1:36:28
um so that and if you have rusted I mean you can't simply copy past C is it I
1:36:31
you can't simply copy past C is it I
1:36:31
you can't simply copy past C is it I mean it's it's a different language
1:36:34
mean it's it's a different language
1:36:34
mean it's it's a different language the team is secretly writing about to
1:36:37
the team is secretly writing about to
1:36:37
the team is secretly writing about to just like rewrite the whole side of this
1:36:38
just like rewrite the whole side of this
1:36:38
just like rewrite the whole side of this extension into rust like over the summer
1:36:40
extension into rust like over the summer
1:36:40
extension into rust like over the summer or something like I'm
1:36:41
or something like I'm
1:36:41
or something like I'm I I don't I don't I don't expect that to
1:36:44
I I don't I don't I don't expect that to
1:36:44
I I don't I don't I don't expect that to happen very attention but uh but but
1:36:46
happen very attention but uh but but
1:36:46
happen very attention but uh but but we're definitely investigating other
1:36:48
we're definitely investigating other
1:36:48
we're definitely investigating other like we're definitely using rust for
1:36:49
like we're definitely using rust for
1:36:49
like we're definitely using rust for some other extensions
1:36:51
some other extensions
1:36:51
some other extensions well thank you thank you so much yelta
1:36:55
well thank you thank you so much yelta
1:36:55
well thank you thank you so much yelta um I know Rob Rob wanted to ask
1:36:57
um I know Rob Rob wanted to ask
1:36:57
um I know Rob Rob wanted to ask questions about the pronunciation of
1:36:58
questions about the pronunciation of
1:36:58
questions about the pronunciation of postgres
1:36:59
postgres
1:36:59
postgres yeah well I just wonder can you get Chad
1:37:02
yeah well I just wonder can you get Chad
1:37:02
yeah well I just wonder can you get Chad gbt to actually call it postgres instead
1:37:04
gbt to actually call it postgres instead
1:37:04
gbt to actually call it postgres instead of postgresql I have not tried but I
1:37:08
of postgresql I have not tried but I
1:37:08
of postgresql I have not tried but I think if you ask it sweet enough I think
1:37:10
think if you ask it sweet enough I think
1:37:10
think if you ask it sweet enough I think it might uh that's that's kind of
1:37:15
so sometimes it has its own free will
1:37:18
so sometimes it has its own free will
1:37:18
so sometimes it has its own free will and that's something that you don't want
1:37:19
and that's something that you don't want
1:37:19
and that's something that you don't want anyway so that's what I've kind of
1:37:21
anyway so that's what I've kind of
1:37:21
anyway so that's what I've kind of learned because it's even if you tell it
1:37:23
learned because it's even if you tell it
1:37:23
learned because it's even if you tell it like oh you don't only want the sequel
1:37:25
like oh you don't only want the sequel
1:37:25
like oh you don't only want the sequel the raw sequel sometimes it still apps
1:37:26
the raw sequel sometimes it still apps
1:37:26
the raw sequel sometimes it still apps like this is the sequel right in front
1:37:28
like this is the sequel right in front
1:37:28
like this is the sequel right in front of it and then and then when it you
1:37:30
of it and then and then when it you
1:37:30
of it and then and then when it you execute it it's it just doesn't work
1:37:31
execute it it's it just doesn't work
1:37:31
execute it it's it just doesn't work because it's I mean that's that's not
1:37:33
because it's I mean that's that's not
1:37:33
because it's I mean that's that's not the sequel that you want to execute
1:37:35
the sequel that you want to execute
1:37:35
the sequel that you want to execute well Rob when we look at our checklist
1:37:37
well Rob when we look at our checklist
1:37:37
well Rob when we look at our checklist we can check off like discussion about
1:37:38
we can check off like discussion about
1:37:39
we can check off like discussion about the pronunciation of postgres because
1:37:40
the pronunciation of postgres because
1:37:40
the pronunciation of postgres because you can't have a postgres conference
1:37:42
you can't have a postgres conference
1:37:42
you can't have a postgres conference without having that discussion
1:37:44
without having that discussion
1:37:44
without having that discussion um so there somebody is working on their
1:37:46
um so there somebody is working on their
1:37:46
um so there somebody is working on their bingo card right now so yeah
1:37:49
bingo card right now so yeah
1:37:49
bingo card right now so yeah you are going to co-host in the emea
1:37:52
you are going to co-host in the emea
1:37:52
you are going to co-host in the emea live stream so for people who come to
1:37:54
live stream so for people who come to
1:37:54
live stream so for people who come to that
1:37:55
that
1:37:55
that um we will see you again
1:37:57
um we will see you again
1:37:57
um we will see you again um tomorrow and for those of you who
1:37:59
um tomorrow and for those of you who
1:37:59
um tomorrow and for those of you who want to mark your calendar
1:38:01
want to mark your calendar
1:38:01
want to mark your calendar um that's happening Wednesday 9cest and
1:38:03
um that's happening Wednesday 9cest and
1:38:03
um that's happening Wednesday 9cest and there's a calendar invite URL showing on
1:38:06
there's a calendar invite URL showing on
1:38:06
there's a calendar invite URL showing on the screen right now so thank you this
1:38:08
the screen right now so thank you this
1:38:08
the screen right now so thank you this is awesome see you tomorrow
1:38:11
interstitial video next is that what's
1:38:13
interstitial video next is that what's
1:38:13
interstitial video next is that what's going on I believe we do uh before we
1:38:16
going on I believe we do uh before we
1:38:16
going on I believe we do uh before we jump to the next speaker uh we wanted to
1:38:18
jump to the next speaker uh we wanted to
1:38:18
jump to the next speaker uh we wanted to kind of highlight some of the on-demand
1:38:19
kind of highlight some of the on-demand
1:38:19
kind of highlight some of the on-demand talks that are available uh and there
1:38:22
talks that are available uh and there
1:38:22
talks that are available uh and there are 25 of them in total this will be
1:38:24
are 25 of them in total this will be
1:38:24
are 25 of them in total this will be just a preview of a few of them uh but
1:38:26
just a preview of a few of them uh but
1:38:27
just a preview of a few of them uh but so hopefully gives people an idea of
1:38:28
so hopefully gives people an idea of
1:38:28
so hopefully gives people an idea of what else is out there uh who haven't
1:38:30
what else is out there uh who haven't
1:38:30
what else is out there uh who haven't you know gone and read all of the guide
1:38:32
you know gone and read all of the guide
1:38:32
you know gone and read all of the guide yet but uh here's a little preview
1:38:36
we are so happy that so many of you are
1:38:39
we are so happy that so many of you are
1:38:39
we are so happy that so many of you are joining us for cytuscon an event for
1:38:41
joining us for cytuscon an event for
1:38:42
joining us for cytuscon an event for postgres 2023 now in its second year I
1:38:45
postgres 2023 now in its second year I
1:38:45
postgres 2023 now in its second year I want to make sure you know there are 37
1:38:47
want to make sure you know there are 37
1:38:48
want to make sure you know there are 37 talks in this year's lineup not just the
1:38:50
talks in this year's lineup not just the
1:38:50
talks in this year's lineup not just the six talks in the two live streams but
1:38:53
six talks in the two live streams but
1:38:54
six talks in the two live streams but there are 25 more brand new talks we
1:38:56
there are 25 more brand new talks we
1:38:56
there are 25 more brand new talks we call them on-demand talks and they're
1:38:58
call them on-demand talks and they're
1:38:58
call them on-demand talks and they're going to publish on YouTube at the very
1:39:00
going to publish on YouTube at the very
1:39:00
going to publish on YouTube at the very start of cytuscon Lucas Borges is giving
1:39:04
start of cytuscon Lucas Borges is giving
1:39:04
start of cytuscon Lucas Borges is giving a demo about how to auto scale Azure
1:39:06
a demo about how to auto scale Azure
1:39:06
a demo about how to auto scale Azure Cosmos DB for postgres using grafana and
1:39:10
Cosmos DB for postgres using grafana and
1:39:10
Cosmos DB for postgres using grafana and Azure serverless hockey Bonita is giving
1:39:13
Azure serverless hockey Bonita is giving
1:39:13
Azure serverless hockey Bonita is giving a wonderful talk on unconventional ways
1:39:15
a wonderful talk on unconventional ways
1:39:15
a wonderful talk on unconventional ways to index uuids in postgres Hedy
1:39:19
to index uuids in postgres Hedy
1:39:19
to index uuids in postgres Hedy dombrovska is presenting on temporal
1:39:21
dombrovska is presenting on temporal
1:39:21
dombrovska is presenting on temporal features and time travel ADI kumara chat
1:39:24
features and time travel ADI kumara chat
1:39:24
features and time travel ADI kumara chat is going to cover partitioning
1:39:26
is going to cover partitioning
1:39:26
is going to cover partitioning strategies for Oracle to postgres
1:39:28
strategies for Oracle to postgres
1:39:28
strategies for Oracle to postgres migrations Bruce momjian is giving one
1:39:31
migrations Bruce momjian is giving one
1:39:31
migrations Bruce momjian is giving one of his many talks on artificial
1:39:33
of his many talks on artificial
1:39:33
of his many talks on artificial intelligence and postgres Paolo
1:39:35
intelligence and postgres Paolo
1:39:35
intelligence and postgres Paolo melchione is going to teach her how to
1:39:37
melchione is going to teach her how to
1:39:37
melchione is going to teach her how to build a web map using Django and post
1:39:40
build a web map using Django and post
1:39:40
build a web map using Django and post GIS and Chelsea Dole has a great talk on
1:39:43
GIS and Chelsea Dole has a great talk on
1:39:43
GIS and Chelsea Dole has a great talk on postgres table bloat and there's even
1:39:45
postgres table bloat and there's even
1:39:45
postgres table bloat and there's even more um you can find all 25 of the
1:39:47
more um you can find all 25 of the
1:39:47
more um you can find all 25 of the on-demand talks that aka.ms slash
1:39:51
on-demand talks that aka.ms slash
1:39:51
on-demand talks that aka.ms slash cytuscon hyphen on demand
1:39:57
all right awesome stuff let's dive in I
1:40:01
all right awesome stuff let's dive in I
1:40:01
all right awesome stuff let's dive in I know that we we spent too much time
1:40:03
know that we we spent too much time
1:40:03
know that we we spent too much time asking questions of Yalta so uh I think
1:40:06
asking questions of Yalta so uh I think
1:40:06
asking questions of Yalta so uh I think we should bring our next speaker on
1:40:09
we should bring our next speaker on
1:40:09
we should bring our next speaker on yeah let's go ahead and bring in Yvonne
1:40:11
yeah let's go ahead and bring in Yvonne
1:40:11
yeah let's go ahead and bring in Yvonne vasmatinov uh who is Hello uh technical
1:40:14
vasmatinov uh who is Hello uh technical
1:40:14
vasmatinov uh who is Hello uh technical lead and accidental DBA how you doing
1:40:17
lead and accidental DBA how you doing
1:40:17
lead and accidental DBA how you doing perfectly fine thanks awesome uh he also
1:40:20
perfectly fine thanks awesome uh he also
1:40:20
perfectly fine thanks awesome uh he also has cytus open source user uh and is
1:40:22
has cytus open source user uh and is
1:40:22
has cytus open source user uh and is going to be talking today about uh Json
1:40:25
going to be talking today about uh Json
1:40:25
going to be talking today about uh Json and analytics and and putting all that
1:40:28
and analytics and and putting all that
1:40:28
and analytics and and putting all that together so let's turn it over to Yvonne
1:40:30
together so let's turn it over to Yvonne
1:40:30
together so let's turn it over to Yvonne and we'll get started
1:40:33
and we'll get started
1:40:33
and we'll get started yes hello everyone uh this talk will
1:40:36
yes hello everyone uh this talk will
1:40:36
yes hello everyone uh this talk will show you a success successful example of
1:40:39
show you a success successful example of
1:40:39
show you a success successful example of real-time analytics based on civil
1:40:42
real-time analytics based on civil
1:40:42
real-time analytics based on civil structured data just on and how cytos
1:40:45
structured data just on and how cytos
1:40:45
structured data just on and how cytos helps with uh shortcomings of that it
1:40:48
helps with uh shortcomings of that it
1:40:48
helps with uh shortcomings of that it will be especially useful for those who
1:40:50
will be especially useful for those who
1:40:51
will be especially useful for those who considering Json B uh for such uh
1:40:54
considering Json B uh for such uh
1:40:54
considering Json B uh for such uh workload to know caveats tips and tricks
1:40:57
workload to know caveats tips and tricks
1:40:57
workload to know caveats tips and tricks and again how status help you with
1:41:01
and again how status help you with
1:41:01
and again how status help you with distribution capabilities
1:41:04
distribution capabilities
1:41:04
distribution capabilities no
1:41:06
no
1:41:06
no uh here is my uh short info and my
1:41:10
uh here is my uh short info and my
1:41:10
uh here is my uh short info and my socials uh we are a company that makes
1:41:13
socials uh we are a company that makes
1:41:13
socials uh we are a company that makes uh social and mobile games you might
1:41:17
uh social and mobile games you might
1:41:17
uh social and mobile games you might have heard about contact Adventures
1:41:19
have heard about contact Adventures
1:41:19
have heard about contact Adventures that's us uh there's also a short list
1:41:22
that's us uh there's also a short list
1:41:22
that's us uh there's also a short list of shortlist of our achievements and we
1:41:26
of shortlist of our achievements and we
1:41:26
of shortlist of our achievements and we are also big Believers in open source
1:41:28
are also big Believers in open source
1:41:28
are also big Believers in open source and encourage every user of Open Source
1:41:30
and encourage every user of Open Source
1:41:30
and encourage every user of Open Source to join the contribution community
1:41:33
to join the contribution community
1:41:33
to join the contribution community now with that out of the way let's get
1:41:35
now with that out of the way let's get
1:41:36
now with that out of the way let's get started so first I'd like to discuss the
1:41:39
started so first I'd like to discuss the
1:41:39
started so first I'd like to discuss the division of labor in our company here we
1:41:42
division of labor in our company here we
1:41:42
division of labor in our company here we have my department internal tools which
1:41:45
have my department internal tools which
1:41:45
have my department internal tools which is essentially a data engineering
1:41:47
is essentially a data engineering
1:41:47
is essentially a data engineering department so we extract data from
1:41:50
department so we extract data from
1:41:50
department so we extract data from various sources we prepare clean it and
1:41:52
various sources we prepare clean it and
1:41:52
various sources we prepare clean it and start into a data warehouse from the
1:41:55
start into a data warehouse from the
1:41:55
start into a data warehouse from the other side there is analytical
1:41:57
other side there is analytical
1:41:57
other side there is analytical Department that uses data we prepared
1:42:00
Department that uses data we prepared
1:42:00
Department that uses data we prepared from reports machine learning tasks Etc
1:42:03
from reports machine learning tasks Etc
1:42:03
from reports machine learning tasks Etc and this talk will mostly focus on the
1:42:07
and this talk will mostly focus on the
1:42:07
and this talk will mostly focus on the data engineering stuff
1:42:10
data engineering stuff
1:42:10
data engineering stuff now to the to our analytical solution
1:42:12
now to the to our analytical solution
1:42:12
now to the to our analytical solution right the general wipe of our solution
1:42:16
right the general wipe of our solution
1:42:16
right the general wipe of our solution can be described as ingest everything
1:42:18
can be described as ingest everything
1:42:18
can be described as ingest everything every piece of data as fast as possible
1:42:22
every piece of data as fast as possible
1:42:22
every piece of data as fast as possible and
1:42:24
and
1:42:24
and with any computational power available
1:42:28
with any computational power available
1:42:28
with any computational power available uh this comes from the following set of
1:42:31
uh this comes from the following set of
1:42:31
uh this comes from the following set of initial requirements from the start we
1:42:35
initial requirements from the start we
1:42:35
initial requirements from the start we were required to have a
1:42:37
were required to have a
1:42:37
were required to have a solutions that capable of storing Sim
1:42:39
solutions that capable of storing Sim
1:42:39
solutions that capable of storing Sim structured data because we as you may
1:42:42
structured data because we as you may
1:42:42
structured data because we as you may have noticed have a lot of games and
1:42:44
have noticed have a lot of games and
1:42:44
have noticed have a lot of games and each game uh has its own unique
1:42:48
each game uh has its own unique
1:42:48
each game uh has its own unique analytical events this unit contains
1:42:52
analytical events this unit contains
1:42:52
analytical events this unit contains content per user so the solution should
1:42:57
content per user so the solution should
1:42:57
content per user so the solution should accept that data at real time it should
1:43:00
accept that data at real time it should
1:43:00
accept that data at real time it should be scalable to support increased demand
1:43:03
be scalable to support increased demand
1:43:03
be scalable to support increased demand if any it should store that
1:43:05
if any it should store that
1:43:05
if any it should store that semi-structured data of users
1:43:07
semi-structured data of users
1:43:07
semi-structured data of users efficiently and allow to efficiently
1:43:10
efficiently and allow to efficiently
1:43:10
efficiently and allow to efficiently query it via SQL and now cytus comes to
1:43:15
query it via SQL and now cytus comes to
1:43:15
query it via SQL and now cytus comes to play but you may wonder why exactly
1:43:18
play but you may wonder why exactly
1:43:18
play but you may wonder why exactly cytos and not time scale DB or click
1:43:21
cytos and not time scale DB or click
1:43:21
cytos and not time scale DB or click house or you go by DB cockroach DB or
1:43:24
house or you go by DB cockroach DB or
1:43:24
house or you go by DB cockroach DB or any other available technology on the
1:43:27
any other available technology on the
1:43:27
any other available technology on the market well along with answer you with
1:43:29
market well along with answer you with
1:43:29
market well along with answer you with the list of pros and cons but it will be
1:43:33
the list of pros and cons but it will be
1:43:33
the list of pros and cons but it will be a little bit dated because we were
1:43:35
a little bit dated because we were
1:43:35
a little bit dated because we were making our choice six years ago so now
1:43:39
making our choice six years ago so now
1:43:39
making our choice six years ago so now status is open source technology
1:43:42
status is open source technology
1:43:42
status is open source technology especially now it supports just be
1:43:47
especially now it supports just be
1:43:47
especially now it supports just be natively because it is an extension of
1:43:49
natively because it is an extension of
1:43:49
natively because it is an extension of postgres this is just on these are
1:43:52
postgres this is just on these are
1:43:52
postgres this is just on these are conveniently queried with SQL familiar
1:43:55
conveniently queried with SQL familiar
1:43:55
conveniently queried with SQL familiar to our users and the set itself is a lab
1:43:59
to our users and the set itself is a lab
1:43:59
to our users and the set itself is a lab Centric extension so it was built with
1:44:01
Centric extension so it was built with
1:44:01
Centric extension so it was built with olap for use case in mind
1:44:04
olap for use case in mind
1:44:04
olap for use case in mind from the other hand there were no Corner
1:44:07
from the other hand there were no Corner
1:44:07
from the other hand there were no Corner storage back then typical for roll up
1:44:10
storage back then typical for roll up
1:44:10
storage back then typical for roll up and there were no convenient way to
1:44:14
and there were no convenient way to
1:44:14
and there were no convenient way to scale the cluster in open source Edition
1:44:17
scale the cluster in open source Edition
1:44:17
scale the cluster in open source Edition it was behind the playable luckily today
1:44:20
it was behind the playable luckily today
1:44:20
it was behind the playable luckily today is no day and both of those issues are
1:44:25
is no day and both of those issues are
1:44:25
is no day and both of those issues are resolved they are an open sourced and
1:44:29
resolved they are an open sourced and
1:44:29
resolved they are an open sourced and thanks Microsoft and especially uh
1:44:33
thanks Microsoft and especially uh
1:44:33
thanks Microsoft and especially uh now let's see the hardware that we have
1:44:36
now let's see the hardware that we have
1:44:36
now let's see the hardware that we have for that
1:44:37
for that
1:44:37
for that we use an on-premise installation due to
1:44:41
we use an on-premise installation due to
1:44:41
we use an on-premise installation due to some historical reasons here you can see
1:44:44
some historical reasons here you can see
1:44:44
some historical reasons here you can see our coordinator machine specs and here
1:44:48
our coordinator machine specs and here
1:44:48
our coordinator machine specs and here are working machines spec so that that
1:44:51
are working machines spec so that that
1:44:51
are working machines spec so that that we have for the off now we rely hugely
1:44:56
we have for the off now we rely hugely
1:44:56
we have for the off now we rely hugely on huge pages one gigabyte in size we
1:45:00
on huge pages one gigabyte in size we
1:45:00
on huge pages one gigabyte in size we disabled swap
1:45:04
the hardware that we have
1:45:08
the hardware that we have
1:45:08
the hardware that we have uh we also picked the battery versus our
1:45:11
uh we also picked the battery versus our
1:45:11
uh we also picked the battery versus our file system uh it's a great file system
1:45:14
file system uh it's a great file system
1:45:14
file system uh it's a great file system uh being copy and write it provides us
1:45:17
uh being copy and write it provides us
1:45:17
uh being copy and write it provides us with the ability to make file system
1:45:20
with the ability to make file system
1:45:20
with the ability to make file system level snapshots which are a perfect
1:45:23
level snapshots which are a perfect
1:45:23
level snapshots which are a perfect solution for quick backups but more
1:45:26
solution for quick backups but more
1:45:26
solution for quick backups but more important importantly the it is it is
1:45:29
important importantly the it is it is
1:45:29
important importantly the it is it is capable of compressing data that it
1:45:31
capable of compressing data that it
1:45:31
capable of compressing data that it manages so using better face we gain a
1:45:36
manages so using better face we gain a
1:45:36
manages so using better face we gain a compression ratio six about six to one
1:45:39
compression ratio six about six to one
1:45:39
compression ratio six about six to one uh using the comp size tool
1:45:43
uh using the comp size tool
1:45:43
uh using the comp size tool and that's essentially how we solve the
1:45:47
and that's essentially how we solve the
1:45:47
and that's essentially how we solve the data issue with Json because as you may
1:45:50
data issue with Json because as you may
1:45:50
data issue with Json because as you may know uh on Sim structured data is very
1:45:54
know uh on Sim structured data is very
1:45:54
know uh on Sim structured data is very space demanding it is great it's six to
1:45:57
space demanding it is great it's six to
1:45:57
space demanding it is great it's six to one is great ratio for us but it could
1:45:59
one is great ratio for us but it could
1:46:00
one is great ratio for us but it could have been better if we played with the
1:46:03
have been better if we played with the
1:46:03
have been better if we played with the sex size compilation uh
1:46:08
sex size compilation uh
1:46:08
sex size compilation uh from the beginning but yeah the time has
1:46:11
from the beginning but yeah the time has
1:46:11
from the beginning but yeah the time has passed so now let's see the describe
1:46:15
passed so now let's see the describe
1:46:15
passed so now let's see the describe solution itself uh here's our example of
1:46:20
solution itself uh here's our example of
1:46:20
solution itself uh here's our example of our data that we accept in our system at
1:46:23
our data that we accept in our system at
1:46:23
our data that we accept in our system at the first three fields are essentially a
1:46:26
the first three fields are essentially a
1:46:26
the first three fields are essentially a primary key of an event
1:46:28
primary key of an event
1:46:28
primary key of an event and data fields can contain any Json
1:46:32
and data fields can contain any Json
1:46:32
and data fields can contain any Json that is required for customer
1:46:35
that is required for customer
1:46:35
that is required for customer in order to store that uh in postgres we
1:46:40
in order to store that uh in postgres we
1:46:40
in order to store that uh in postgres we dynamically create tables that
1:46:42
dynamically create tables that
1:46:42
dynamically create tables that correspond to the name and other fields
1:46:45
correspond to the name and other fields
1:46:45
correspond to the name and other fields go as is essentially but there is also a
1:46:50
go as is essentially but there is also a
1:46:50
go as is essentially but there is also a special profile
1:46:52
special profile
1:46:52
special profile events designed to store Common data for
1:46:57
events designed to store Common data for
1:46:57
events designed to store Common data for events so it's like normalization
1:47:00
events so it's like normalization
1:47:00
events so it's like normalization there is a permanent profile which is
1:47:03
there is a permanent profile which is
1:47:03
there is a permanent profile which is actually updatable entity so it is
1:47:06
actually updatable entity so it is
1:47:06
actually updatable entity so it is designed to store data like register
1:47:08
designed to store data like register
1:47:08
designed to store data like register date or country something that doesn't
1:47:10
date or country something that doesn't
1:47:10
date or country something that doesn't change over time for all events and
1:47:13
change over time for all events and
1:47:13
change over time for all events and there is a volatile profile for things
1:47:17
there is a volatile profile for things
1:47:17
there is a volatile profile for things like user level or something it is
1:47:19
like user level or something it is
1:47:19
like user level or something it is similar to events it is not updated it
1:47:21
similar to events it is not updated it
1:47:21
similar to events it is not updated it is stored but upon insertion we
1:47:24
is stored but upon insertion we
1:47:24
is stored but upon insertion we calculate a range of activity for the
1:47:26
calculate a range of activity for the
1:47:26
calculate a range of activity for the profile so the profile considered active
1:47:30
profile so the profile considered active
1:47:30
profile so the profile considered active from the update time where the profile
1:47:32
from the update time where the profile
1:47:32
from the update time where the profile came to our system and up to the next
1:47:35
came to our system and up to the next
1:47:35
came to our system and up to the next profile for that users
1:47:37
profile for that users
1:47:37
profile for that users uh combining all this together we get
1:47:41
uh combining all this together we get
1:47:41
uh combining all this together we get the base for most of our analytical
1:47:44
the base for most of our analytical
1:47:44
the base for most of our analytical queries so this join represents like 95
1:47:48
queries so this join represents like 95
1:47:48
queries so this join represents like 95 of queries that are performed in our
1:47:51
of queries that are performed in our
1:47:51
of queries that are performed in our system
1:47:52
system
1:47:52
system and uh since this per user analytics all
1:47:58
and uh since this per user analytics all
1:47:58
and uh since this per user analytics all of our tables presented here are
1:48:00
of our tables presented here are
1:48:00
of our tables presented here are distributed by the user ID column
1:48:04
distributed by the user ID column
1:48:04
distributed by the user ID column now I also should mention that we have
1:48:06
now I also should mention that we have
1:48:06
now I also should mention that we have more tables to support integration with
1:48:10
more tables to support integration with
1:48:10
more tables to support integration with Partners like cup slider Google Etc but
1:48:14
Partners like cup slider Google Etc but
1:48:14
Partners like cup slider Google Etc but it is a big part of our analytics but it
1:48:17
it is a big part of our analytics but it
1:48:17
it is a big part of our analytics but it is not a part of this topic maybe next
1:48:20
is not a part of this topic maybe next
1:48:20
is not a part of this topic maybe next time the stock sorry maybe next time
1:48:22
time the stock sorry maybe next time
1:48:22
time the stock sorry maybe next time next session so let's see some actual
1:48:26
next session so let's see some actual
1:48:26
next session so let's see some actual results of this uh here you see the
1:48:31
results of this uh here you see the
1:48:31
results of this uh here you see the statistics that we gathered from
1:48:33
statistics that we gathered from
1:48:33
statistics that we gathered from bugister statements extension in about
1:48:36
bugister statements extension in about
1:48:36
bugister statements extension in about two months and as you can see the vast
1:48:40
two months and as you can see the vast
1:48:40
two months and as you can see the vast majority of our analytical queries runs
1:48:43
majority of our analytical queries runs
1:48:43
majority of our analytical queries runs under 30 seconds here is a quick plot to
1:48:48
under 30 seconds here is a quick plot to
1:48:48
under 30 seconds here is a quick plot to demonstrate it visually and now let's
1:48:51
demonstrate it visually and now let's
1:48:51
demonstrate it visually and now let's discuss how we gained it using Json b as
1:48:55
discuss how we gained it using Json b as
1:48:55
discuss how we gained it using Json b as our main type of data
1:48:58
our main type of data
1:48:58
our main type of data uh so there are two caveats that you
1:49:01
uh so there are two caveats that you
1:49:01
uh so there are two caveats that you should be aware when you're using Json B
1:49:03
should be aware when you're using Json B
1:49:03
should be aware when you're using Json B first this Json B's got get toasted
1:49:07
first this Json B's got get toasted
1:49:07
first this Json B's got get toasted because they are an autistic restricted
1:49:09
because they are an autistic restricted
1:49:09
because they are an autistic restricted in size and essentially if they get if
1:49:13
in size and essentially if they get if
1:49:13
in size and essentially if they get if they get toasted uh it will hurt the
1:49:15
they get toasted uh it will hurt the
1:49:15
they get toasted uh it will hurt the performance but secondly and most
1:49:17
performance but secondly and most
1:49:17
performance but secondly and most importantly uh past grass still does not
1:49:20
importantly uh past grass still does not
1:49:20
importantly uh past grass still does not gather planner statistics for Json B
1:49:23
gather planner statistics for Json B
1:49:23
gather planner statistics for Json B content essentially uh what it leads to
1:49:27
content essentially uh what it leads to
1:49:27
content essentially uh what it leads to if you have some query that uses uh
1:49:31
if you have some query that uses uh
1:49:31
if you have some query that uses uh Justin B in where class uh your rows
1:49:36
Justin B in where class uh your rows
1:49:36
Justin B in where class uh your rows gets very underestimated extremely
1:49:38
gets very underestimated extremely
1:49:38
gets very underestimated extremely underestimated and you will get
1:49:41
underestimated and you will get
1:49:41
underestimated and you will get sub-optimal plan so what can we do about
1:49:45
sub-optimal plan so what can we do about
1:49:45
sub-optimal plan so what can we do about that uh well to battle toasts we advise
1:49:49
that uh well to battle toasts we advise
1:49:49
that uh well to battle toasts we advise our clients to use thinner events and
1:49:52
our clients to use thinner events and
1:49:52
our clients to use thinner events and bigger profiles so the vast majority of
1:49:55
bigger profiles so the vast majority of
1:49:55
bigger profiles so the vast majority of events won't get toasted and profiles
1:49:58
events won't get toasted and profiles
1:49:58
events won't get toasted and profiles will be kept in
1:50:00
will be kept in
1:50:00
will be kept in it can go even further with via
1:50:04
it can go even further with via
1:50:04
it can go even further with via restrictions via check constraints or
1:50:07
restrictions via check constraints or
1:50:07
restrictions via check constraints or something on the middleware side but we
1:50:10
something on the middleware side but we
1:50:10
something on the middleware side but we do not recommend it it's too restrictive
1:50:11
do not recommend it it's too restrictive
1:50:12
do not recommend it it's too restrictive and you can play around with table
1:50:14
and you can play around with table
1:50:14
and you can play around with table settings or even calculation setting for
1:50:16
settings or even calculation setting for
1:50:16
settings or even calculation setting for postgres but once again we didn't write
1:50:19
postgres but once again we didn't write
1:50:19
postgres but once again we didn't write we didn't need it so far
1:50:23
we didn't need it so far
1:50:23
we didn't need it so far now uh statistics first what you can do
1:50:26
now uh statistics first what you can do
1:50:26
now uh statistics first what you can do with Statistics is just accept accept
1:50:29
with Statistics is just accept accept
1:50:29
with Statistics is just accept accept the fact that some of your plans will be
1:50:31
the fact that some of your plans will be
1:50:31
the fact that some of your plans will be some optimal and maybe
1:50:34
some optimal and maybe
1:50:34
some optimal and maybe add more computational power to it
1:50:38
add more computational power to it
1:50:38
add more computational power to it but if you need to see uh character
1:50:42
but if you need to see uh character
1:50:42
but if you need to see uh character estimation so you can
1:50:45
estimation so you can
1:50:45
estimation so you can spend some some space on disk and create
1:50:48
spend some some space on disk and create
1:50:48
spend some some space on disk and create a gene Index this is a special type of
1:50:51
a gene Index this is a special type of
1:50:51
a gene Index this is a special type of index specifically designed for
1:50:52
index specifically designed for
1:50:52
index specifically designed for composite values and most importantly
1:50:55
composite values and most importantly
1:50:55
composite values and most importantly for jsons that it supports uh just some
1:50:58
for jsons that it supports uh just some
1:50:58
for jsons that it supports uh just some path expressions
1:50:59
path expressions
1:50:59
path expressions uh here's an example of gen index and
1:51:02
uh here's an example of gen index and
1:51:02
uh here's an example of gen index and Justin path expression and as you can
1:51:06
Justin path expression and as you can
1:51:06
Justin path expression and as you can see uh row estimations got much better
1:51:10
see uh row estimations got much better
1:51:10
see uh row estimations got much better it's a great solution it covers all
1:51:14
it's a great solution it covers all
1:51:14
it's a great solution it covers all fields of the field or columns within it
1:51:18
fields of the field or columns within it
1:51:18
fields of the field or columns within it but unfortunately it is much slower than
1:51:21
but unfortunately it is much slower than
1:51:21
but unfortunately it is much slower than B3 index because it uses bitmap index
1:51:25
B3 index because it uses bitmap index
1:51:25
B3 index because it uses bitmap index guns and just on pass itself while being
1:51:28
guns and just on pass itself while being
1:51:28
guns and just on pass itself while being pretty powerful in capabilities it's not
1:51:31
pretty powerful in capabilities it's not
1:51:31
pretty powerful in capabilities it's not very readable and reminds me of regular
1:51:35
very readable and reminds me of regular
1:51:35
very readable and reminds me of regular expressions
1:51:36
expressions
1:51:36
expressions uh and it can get very ugly very ugly
1:51:40
uh and it can get very ugly very ugly
1:51:40
uh and it can get very ugly very ugly and very fast
1:51:41
and very fast
1:51:42
and very fast so now let's talk about those between
1:51:45
so now let's talk about those between
1:51:45
so now let's talk about those between indexes well that's quite simple you can
1:51:48
indexes well that's quite simple you can
1:51:48
indexes well that's quite simple you can create them on this Arrow operator and
1:51:52
create them on this Arrow operator and
1:51:52
create them on this Arrow operator and this bit index present the performance
1:51:54
this bit index present the performance
1:51:54
this bit index present the performance is the best the most emissions is the
1:51:59
is the best the most emissions is the
1:51:59
is the best the most emissions is the most accurate Etc but you'll have to
1:52:01
most accurate Etc but you'll have to
1:52:01
most accurate Etc but you'll have to create it on each column of just some
1:52:04
create it on each column of just some
1:52:04
create it on each column of just some separately
1:52:06
separately
1:52:06
separately now there is also an ultimate transverse
1:52:09
now there is also an ultimate transverse
1:52:09
now there is also an ultimate transverse answer for all our
1:52:11
answer for all our
1:52:11
answer for all our performance issues statistics toast Etc
1:52:14
performance issues statistics toast Etc
1:52:14
performance issues statistics toast Etc just add row computational power I mean
1:52:17
just add row computational power I mean
1:52:17
just add row computational power I mean up optimal plan or sub-optimal if you
1:52:20
up optimal plan or sub-optimal if you
1:52:21
up optimal plan or sub-optimal if you have enough computational power your
1:52:23
have enough computational power your
1:52:23
have enough computational power your queries will be will perform adequately
1:52:26
queries will be will perform adequately
1:52:26
queries will be will perform adequately and of course it's mostly joke it's once
1:52:30
and of course it's mostly joke it's once
1:52:30
and of course it's mostly joke it's once off all your problems but again the
1:52:32
off all your problems but again the
1:52:32
off all your problems but again the status it's a very great option to have
1:52:36
status it's a very great option to have
1:52:36
status it's a very great option to have sometimes
1:52:37
sometimes
1:52:37
sometimes uh now we also have some default indexes
1:52:40
uh now we also have some default indexes
1:52:40
uh now we also have some default indexes that we create for all games to support
1:52:42
that we create for all games to support
1:52:42
that we create for all games to support our base join right but we also
1:52:46
our base join right but we also
1:52:46
our base join right but we also sometimes need to resort to the
1:52:47
sometimes need to resort to the
1:52:47
sometimes need to resort to the normalization because our data can get
1:52:51
normalization because our data can get
1:52:51
normalization because our data can get very big for example our biggest table
1:52:53
very big for example our biggest table
1:52:53
very big for example our biggest table is 28 terabytes in size and indexes
1:52:57
is 28 terabytes in size and indexes
1:52:57
is 28 terabytes in size and indexes indexes this join these profiles will
1:53:00
indexes this join these profiles will
1:53:00
indexes this join these profiles will not perform well so what we do instead
1:53:03
not perform well so what we do instead
1:53:03
not perform well so what we do instead is uh just store that value in the
1:53:07
is uh just store that value in the
1:53:07
is uh just store that value in the event table itself in order to do that
1:53:10
event table itself in order to do that
1:53:10
event table itself in order to do that we just create a cross pointing column
1:53:13
we just create a cross pointing column
1:53:13
we just create a cross pointing column uh write a simple update queries that
1:53:16
uh write a simple update queries that
1:53:16
uh write a simple update queries that fuels that column using the select you
1:53:19
fuels that column using the select you
1:53:19
fuels that column using the select you saw on the slide uh this update is
1:53:22
saw on the slide uh this update is
1:53:22
saw on the slide uh this update is wrapped up into callable nice callable
1:53:24
wrapped up into callable nice callable
1:53:24
wrapped up into callable nice callable procedure and the closest procedure is
1:53:27
procedure and the closest procedure is
1:53:27
procedure and the closest procedure is scaled scheduled when it is appropriate
1:53:30
scaled scheduled when it is appropriate
1:53:30
scaled scheduled when it is appropriate for us using this approach we eliminate
1:53:33
for us using this approach we eliminate
1:53:33
for us using this approach we eliminate the needs of this slow join and it gets
1:53:38
the needs of this slow join and it gets
1:53:38
the needs of this slow join and it gets great results but from the other hand uh
1:53:42
great results but from the other hand uh
1:53:42
great results but from the other hand uh uh now we have some historical databases
1:53:47
uh now we have some historical databases
1:53:47
uh now we have some historical databases fill this column field and fresh data
1:53:50
fill this column field and fresh data
1:53:50
fill this column field and fresh data where it is not filled no problem if you
1:53:54
where it is not filled no problem if you
1:53:54
where it is not filled no problem if you know the threshold of filling you can
1:53:56
know the threshold of filling you can
1:53:56
know the threshold of filling you can just Union fresh data and historical
1:53:58
just Union fresh data and historical
1:53:58
just Union fresh data and historical data
1:54:00
data
1:54:00
data it is similar to mod views
1:54:04
it is similar to mod views
1:54:04
it is similar to mod views similar to what use might be and let's
1:54:07
similar to what use might be and let's
1:54:07
similar to what use might be and let's see how status
1:54:10
see how status
1:54:10
see how status so uh here is a typical analytical query
1:54:13
so uh here is a typical analytical query
1:54:13
so uh here is a typical analytical query and when we want to store it on disk we
1:54:17
and when we want to store it on disk we
1:54:17
and when we want to store it on disk we just create mod view and schedule its
1:54:19
just create mod view and schedule its
1:54:19
just create mod view and schedule its refresh right so simple all okay now the
1:54:23
refresh right so simple all okay now the
1:54:23
refresh right so simple all okay now the status you just create mod view invoke
1:54:26
status you just create mod view invoke
1:54:26
status you just create mod view invoke redistributed table on that mod view and
1:54:29
redistributed table on that mod view and
1:54:29
redistributed table on that mod view and schedule its refresh now it's joinable
1:54:32
schedule its refresh now it's joinable
1:54:32
schedule its refresh now it's joinable with all other distributed data in our
1:54:35
with all other distributed data in our
1:54:35
with all other distributed data in our cluster right nope unfortunately there
1:54:39
cluster right nope unfortunately there
1:54:39
cluster right nope unfortunately there is no way to create distributed module
1:54:41
is no way to create distributed module
1:54:41
is no way to create distributed module in status and it is by Design because
1:54:44
in status and it is by Design because
1:54:44
in status and it is by Design because developers suggest us to use inserting
1:54:47
developers suggest us to use inserting
1:54:47
developers suggest us to use inserting to select functionality
1:54:50
to select functionality
1:54:50
to select functionality instead no problem in order to do that
1:54:53
instead no problem in order to do that
1:54:53
instead no problem in order to do that just create table
1:54:55
just create table
1:54:55
just create table regular table distributed and right
1:54:59
regular table distributed and right
1:54:59
regular table distributed and right truncate insert procedure that mimics
1:55:03
truncate insert procedure that mimics
1:55:03
truncate insert procedure that mimics essentially
1:55:05
essentially
1:55:05
essentially materialized view functionality
1:55:08
materialized view functionality
1:55:08
materialized view functionality now all right uh that's the gist of our
1:55:13
now all right uh that's the gist of our
1:55:13
now all right uh that's the gist of our solution now let's see how we perform
1:55:15
solution now let's see how we perform
1:55:15
solution now let's see how we perform maintenance of all that beautiful beasts
1:55:17
maintenance of all that beautiful beasts
1:55:17
maintenance of all that beautiful beasts so our maintenance is essentially split
1:55:20
so our maintenance is essentially split
1:55:20
so our maintenance is essentially split between the developers who write uh uh
1:55:24
between the developers who write uh uh
1:55:24
between the developers who write uh uh interesting middleware because uh some
1:55:28
interesting middleware because uh some
1:55:28
interesting middleware because uh some tables just a part of essentially a part
1:55:30
tables just a part of essentially a part
1:55:30
tables just a part of essentially a part of that middleware uh but it is wrapped
1:55:34
of that middleware uh but it is wrapped
1:55:34
of that middleware uh but it is wrapped into DBA maintenance that creates
1:55:38
into DBA maintenance that creates
1:55:38
into DBA maintenance that creates infrastructure databases
1:55:40
infrastructure databases
1:55:40
infrastructure databases Etc roles for that middleware but from
1:55:43
Etc roles for that middleware but from
1:55:43
Etc roles for that middleware but from the other side it creates views and
1:55:45
the other side it creates views and
1:55:45
the other side it creates views and indexes on tables that are managed by
1:55:47
indexes on tables that are managed by
1:55:47
indexes on tables that are managed by middleware but either way uh both parts
1:55:51
middleware but either way uh both parts
1:55:51
middleware but either way uh both parts are managed by a flyweight database and
1:55:53
are managed by a flyweight database and
1:55:53
are managed by a flyweight database and database migration tool which allows to
1:55:56
database migration tool which allows to
1:55:56
database migration tool which allows to store changes to the base in a sequence
1:55:58
store changes to the base in a sequence
1:55:58
store changes to the base in a sequence of uh version migration Scripts
1:56:03
of uh version migration Scripts
1:56:03
of uh version migration Scripts uh now now rdb is use follow the
1:56:07
uh now now rdb is use follow the
1:56:07
uh now now rdb is use follow the github's approach it means that we have
1:56:10
github's approach it means that we have
1:56:10
github's approach it means that we have a single repository with automated
1:56:12
a single repository with automated
1:56:12
a single repository with automated delivery configured and every change to
1:56:15
delivery configured and every change to
1:56:15
delivery configured and every change to the database is performed as a commute
1:56:17
the database is performed as a commute
1:56:17
the database is performed as a commute to the repository with a
1:56:20
to the repository with a
1:56:20
to the repository with a migration script
1:56:22
migration script
1:56:22
migration script uh very convenience of this in lock Etc
1:56:26
uh very convenience of this in lock Etc
1:56:26
uh very convenience of this in lock Etc now let's see how we handle
1:56:28
now let's see how we handle
1:56:28
now let's see how we handle multi-tenancy because we have a lot of
1:56:30
multi-tenancy because we have a lot of
1:56:30
multi-tenancy because we have a lot of games each game is essentially a tenant
1:56:33
games each game is essentially a tenant
1:56:33
games each game is essentially a tenant so there are preferred ways to do a
1:56:35
so there are preferred ways to do a
1:56:35
so there are preferred ways to do a multi-tenancy in status uh uh there is a
1:56:39
multi-tenancy in status uh uh there is a
1:56:39
multi-tenancy in status uh uh there is a way where you have a tenant column in
1:56:41
way where you have a tenant column in
1:56:41
way where you have a tenant column in all tables uh all distributed tables it
1:56:45
all tables uh all distributed tables it
1:56:45
all tables uh all distributed tables it is the way described in the
1:56:46
is the way described in the
1:56:46
is the way described in the documentation you can also have a schema
1:56:49
documentation you can also have a schema
1:56:49
documentation you can also have a schema schema per tenant in your database we
1:56:54
schema per tenant in your database we
1:56:54
schema per tenant in your database we went another way and created a database
1:56:57
went another way and created a database
1:56:57
went another way and created a database based multi-tennessee so we'll create a
1:56:59
based multi-tennessee so we'll create a
1:57:00
based multi-tennessee so we'll create a database per tenant because we like that
1:57:02
database per tenant because we like that
1:57:02
database per tenant because we like that strong isolation between tenant uh we
1:57:05
strong isolation between tenant uh we
1:57:05
strong isolation between tenant uh we like that we can have a custom
1:57:07
like that we can have a custom
1:57:07
like that we can have a custom configuration of database pertainant
1:57:09
configuration of database pertainant
1:57:09
configuration of database pertainant especially
1:57:11
especially
1:57:11
especially cytos configuration so we can have
1:57:13
cytos configuration so we can have
1:57:13
cytos configuration so we can have different workers for different
1:57:15
different workers for different
1:57:15
different workers for different databases it was not so easy but uh five
1:57:20
databases it was not so easy but uh five
1:57:20
databases it was not so easy but uh five or six years ago yeah and surprisingly
1:57:23
or six years ago yeah and surprisingly
1:57:23
or six years ago yeah and surprisingly to us it is much easier to maintain uh
1:57:27
to us it is much easier to maintain uh
1:57:27
to us it is much easier to maintain uh multiple tenants this way because we can
1:57:29
multiple tenants this way because we can
1:57:29
multiple tenants this way because we can invoke for example parallel migrations
1:57:31
invoke for example parallel migrations
1:57:31
invoke for example parallel migrations for some common or similar objects a bit
1:57:35
for some common or similar objects a bit
1:57:35
for some common or similar objects a bit alternates from the other hand there are
1:57:37
alternates from the other hand there are
1:57:37
alternates from the other hand there are no as you may know there are no cross
1:57:39
no as you may know there are no cross
1:57:39
no as you may know there are no cross database queries in postgres itself
1:57:42
database queries in postgres itself
1:57:42
database queries in postgres itself I'm more than sure that every podcast
1:57:44
I'm more than sure that every podcast
1:57:44
I'm more than sure that every podcast user have seen this server
1:57:47
user have seen this server
1:57:47
user have seen this server and uh yeah we'll work around it using
1:57:50
and uh yeah we'll work around it using
1:57:50
and uh yeah we'll work around it using the building extension usual story here
1:57:54
the building extension usual story here
1:57:54
the building extension usual story here but more importantly uh approach is
1:57:57
but more importantly uh approach is
1:57:57
but more importantly uh approach is poorly supported by cytos itself because
1:58:00
poorly supported by cytos itself because
1:58:00
poorly supported by cytos itself because one of the best we said this requires uh
1:58:03
one of the best we said this requires uh
1:58:03
one of the best we said this requires uh the demon process on both machines it
1:58:07
the demon process on both machines it
1:58:07
the demon process on both machines it requires and connections for any number
1:58:09
requires and connections for any number
1:58:09
requires and connections for any number of workers in cluster for deadlock
1:58:11
of workers in cluster for deadlock
1:58:11
of workers in cluster for deadlock detections adduction and two connections
1:58:14
detections adduction and two connections
1:58:14
detections adduction and two connections for uh transaction recovery purposes
1:58:19
for uh transaction recovery purposes
1:58:19
for uh transaction recovery purposes no let's see let's assume that we have a
1:58:23
no let's see let's assume that we have a
1:58:23
no let's see let's assume that we have a cluster with 40 machines and 30
1:58:25
cluster with 40 machines and 30
1:58:25
cluster with 40 machines and 30 databases on these machines workers
1:58:27
databases on these machines workers
1:58:27
databases on these machines workers doing a little math we get 1260
1:58:31
doing a little math we get 1260
1:58:31
doing a little math we get 1260 connections only for maintenance
1:58:32
connections only for maintenance
1:58:32
connections only for maintenance purposes per worker uh 60 of them are
1:58:36
purposes per worker uh 60 of them are
1:58:36
purposes per worker uh 60 of them are required each minute by default and will
1:58:40
required each minute by default and will
1:58:40
required each minute by default and will hundred I just stay idle all the time uh
1:58:44
hundred I just stay idle all the time uh
1:58:44
hundred I just stay idle all the time uh what did we do what did we do to battle
1:58:46
what did we do what did we do to battle
1:58:46
what did we do what did we do to battle that is disabled automatic recovery and
1:58:49
that is disabled automatic recovery and
1:58:49
that is disabled automatic recovery and Deadlock detection and schedule it
1:58:51
Deadlock detection and schedule it
1:58:51
Deadlock detection and schedule it sequential location on all databases
1:58:55
sequential location on all databases
1:58:55
sequential location on all databases ourselves luckily tutus provide us with
1:58:58
ourselves luckily tutus provide us with
1:58:58
ourselves luckily tutus provide us with the functions to invoke both of that
1:59:01
the functions to invoke both of that
1:59:01
the functions to invoke both of that manually
1:59:02
manually
1:59:02
manually but it will not solve all your problems
1:59:04
but it will not solve all your problems
1:59:04
but it will not solve all your problems there are still issues with that
1:59:07
there are still issues with that
1:59:07
there are still issues with that another point of playing in our
1:59:09
another point of playing in our
1:59:09
another point of playing in our maintenance process is multi-level mod
1:59:11
maintenance process is multi-level mod
1:59:11
maintenance process is multi-level mod views so consider a scenario where you
1:59:14
views so consider a scenario where you
1:59:14
views so consider a scenario where you have uh mod view that depends on another
1:59:17
have uh mod view that depends on another
1:59:17
have uh mod view that depends on another mod views and you will have to obviously
1:59:20
mod views and you will have to obviously
1:59:20
mod views and you will have to obviously to schedule the refreshes in a strict
1:59:23
to schedule the refreshes in a strict
1:59:23
to schedule the refreshes in a strict succession
1:59:24
succession
1:59:24
succession what's wrong with that well you will
1:59:26
what's wrong with that well you will
1:59:26
what's wrong with that well you will have you'll have to pick uh time of
1:59:29
have you'll have to pick uh time of
1:59:29
have you'll have to pick uh time of refresh carefully and manually and you
1:59:33
refresh carefully and manually and you
1:59:33
refresh carefully and manually and you are at risk to get still data if there
1:59:35
are at risk to get still data if there
1:59:35
are at risk to get still data if there is a delays of Refreshers somewhere in
1:59:37
is a delays of Refreshers somewhere in
1:59:37
is a delays of Refreshers somewhere in the chain
1:59:38
the chain
1:59:38
the chain this is a chain so what we did we do to
1:59:41
this is a chain so what we did we do to
1:59:41
this is a chain so what we did we do to control that we implemented a simple
1:59:44
control that we implemented a simple
1:59:44
control that we implemented a simple event-based solution neatly integrated
1:59:47
event-based solution neatly integrated
1:59:47
event-based solution neatly integrated with our ecl middleware to essentially
1:59:50
with our ecl middleware to essentially
1:59:50
with our ecl middleware to essentially configure triggers and now our views are
1:59:54
configure triggers and now our views are
1:59:54
configure triggers and now our views are refreshed on triggers not the best
1:59:57
refreshed on triggers not the best
1:59:57
refreshed on triggers not the best triggers but conceptually triggers when
2:00:00
triggers but conceptually triggers when
2:00:00
triggers but conceptually triggers when their dependents are ready
2:00:03
their dependents are ready
2:00:03
their dependents are ready what we got is get got rid of all Chrome
2:00:06
what we got is get got rid of all Chrome
2:00:06
what we got is get got rid of all Chrome Expressions now all the objects are
2:00:10
Expressions now all the objects are
2:00:10
Expressions now all the objects are updated automatically and we do not get
2:00:14
updated automatically and we do not get
2:00:14
updated automatically and we do not get stale data Refreshers may be delayed but
2:00:17
stale data Refreshers may be delayed but
2:00:17
stale data Refreshers may be delayed but data will always be fresh in those chain
2:00:21
data will always be fresh in those chain
2:00:21
data will always be fresh in those chain and as Knights a nice side effect we get
2:00:25
and as Knights a nice side effect we get
2:00:25
and as Knights a nice side effect we get data to create the fancy uh
2:00:29
data to create the fancy uh
2:00:29
data to create the fancy uh actualization dashboard to see the state
2:00:32
actualization dashboard to see the state
2:00:32
actualization dashboard to see the state of refreshes for our views
2:00:35
of refreshes for our views
2:00:35
of refreshes for our views now let's talk rebalancing here very
2:00:38
now let's talk rebalancing here very
2:00:38
now let's talk rebalancing here very important part for a distributed
2:00:40
important part for a distributed
2:00:40
important part for a distributed database of course I once again remind
2:00:43
database of course I once again remind
2:00:43
database of course I once again remind you that it got much easier in recent
2:00:46
you that it got much easier in recent
2:00:46
you that it got much easier in recent years hooray and thank you Microsoft and
2:00:48
years hooray and thank you Microsoft and
2:00:48
years hooray and thank you Microsoft and cytos but there are some minor issues
2:00:50
cytos but there are some minor issues
2:00:50
cytos but there are some minor issues with that minor first one is well issues
2:00:53
with that minor first one is well issues
2:00:53
with that minor first one is well issues uh not all at skates discovered with
2:00:56
uh not all at skates discovered with
2:00:56
uh not all at skates discovered with that process there may be some issues
2:00:59
that process there may be some issues
2:00:59
that process there may be some issues that you will have to deal with it but
2:01:01
that you will have to deal with it but
2:01:01
that you will have to deal with it but uh
2:01:03
uh
2:01:03
uh more importantly uh uh default balancing
2:01:07
more importantly uh uh default balancing
2:01:07
more importantly uh uh default balancing leads to underutilization of cluster
2:01:09
leads to underutilization of cluster
2:01:09
leads to underutilization of cluster Also let's consider the situation when
2:01:11
Also let's consider the situation when
2:01:11
Also let's consider the situation when we have two workers with four CPUs and
2:01:13
we have two workers with four CPUs and
2:01:13
we have two workers with four CPUs and four shots per worker now we add two new
2:01:18
four shots per worker now we add two new
2:01:18
four shots per worker now we add two new machines to that cluster and invoke a
2:01:20
machines to that cluster and invoke a
2:01:20
machines to that cluster and invoke a rebalance after that we will get two
2:01:23
rebalance after that we will get two
2:01:24
rebalance after that we will get two charts period machine
2:01:25
charts period machine
2:01:25
charts period machine as you may see two is less than four so
2:01:30
as you may see two is less than four so
2:01:30
as you may see two is less than four so to start clearly lead to the
2:01:33
to start clearly lead to the
2:01:33
to start clearly lead to the underutilization of our newly rebalanced
2:01:37
underutilization of our newly rebalanced
2:01:37
underutilization of our newly rebalanced cluster
2:01:38
cluster
2:01:38
cluster instead of rebalancing you can do ultra
2:01:42
instead of rebalancing you can do ultra
2:01:42
instead of rebalancing you can do ultra distributed table and change the short
2:01:44
distributed table and change the short
2:01:44
distributed table and change the short count of the table in question
2:01:47
count of the table in question
2:01:47
count of the table in question uh that way it will it will work as
2:01:50
uh that way it will it will work as
2:01:50
uh that way it will it will work as expected you'll get a nice layout of
2:01:52
expected you'll get a nice layout of
2:01:52
expected you'll get a nice layout of shards according to the number of CPUs
2:01:55
shards according to the number of CPUs
2:01:55
shards according to the number of CPUs but there are some considerable
2:01:57
but there are some considerable
2:01:57
but there are some considerable drawbacks with this approach first there
2:02:00
drawbacks with this approach first there
2:02:00
drawbacks with this approach first there is no convenient start function like
2:02:02
is no convenient start function like
2:02:02
is no convenient start function like this defaulty balancing that was open
2:02:04
this defaulty balancing that was open
2:02:04
this defaulty balancing that was open sourced you will have to alter each
2:02:07
sourced you will have to alter each
2:02:07
sourced you will have to alter each table that you want to alter manually
2:02:10
table that you want to alter manually
2:02:10
table that you want to alter manually by name
2:02:12
by name
2:02:12
by name and what's more important uh when table
2:02:16
and what's more important uh when table
2:02:16
and what's more important uh when table short count is altered it is recreated
2:02:19
short count is altered it is recreated
2:02:19
short count is altered it is recreated so you will get a new table with new
2:02:21
so you will get a new table with new
2:02:21
so you will get a new table with new Orit after alteration and assume
2:02:25
Orit after alteration and assume
2:02:25
Orit after alteration and assume suggest it means that you will have to
2:02:28
suggest it means that you will have to
2:02:28
suggest it means that you will have to recreate all dependent view undependent
2:02:30
recreate all dependent view undependent
2:02:30
recreate all dependent view undependent object views indexes Etc
2:02:33
object views indexes Etc
2:02:33
object views indexes Etc stress to do that but unfortunately
2:02:36
stress to do that but unfortunately
2:02:36
stress to do that but unfortunately there are issues with that as well
2:02:39
there are issues with that as well
2:02:39
there are issues with that as well now at time if you want to know more
2:02:43
now at time if you want to know more
2:02:43
now at time if you want to know more about our issues with faced and how we
2:02:47
about our issues with faced and how we
2:02:47
about our issues with faced and how we our practices and conventions there's a
2:02:49
our practices and conventions there's a
2:02:49
our practices and conventions there's a series of posts on the Zone where I am
2:02:52
series of posts on the Zone where I am
2:02:52
series of posts on the Zone where I am after where I describe that in much more
2:02:55
after where I describe that in much more
2:02:55
after where I describe that in much more detail
2:02:57
detail
2:02:57
detail now let's talk a user experience that we
2:03:00
now let's talk a user experience that we
2:03:00
now let's talk a user experience that we users have with our system let's talk
2:03:03
users have with our system let's talk
2:03:03
users have with our system let's talk expectations we expected our users to
2:03:05
expectations we expected our users to
2:03:05
expectations we expected our users to have frequent queries with considerably
2:03:08
have frequent queries with considerably
2:03:08
have frequent queries with considerably with reasonably small results so all the
2:03:11
with reasonably small results so all the
2:03:11
with reasonably small results so all the substantial aggregation will be
2:03:12
substantial aggregation will be
2:03:12
substantial aggregation will be performed at DB level
2:03:14
performed at DB level
2:03:14
performed at DB level unfortunately in reality our users
2:03:17
unfortunately in reality our users
2:03:17
unfortunately in reality our users perform one query per day with a slow
2:03:20
perform one query per day with a slow
2:03:20
perform one query per day with a slow query which loads like gigabytes data of
2:03:24
query which loads like gigabytes data of
2:03:24
query which loads like gigabytes data of size
2:03:25
size
2:03:25
size why so uh well essentially because bi
2:03:29
why so uh well essentially because bi
2:03:29
why so uh well essentially because bi tools are encourage this so for example
2:03:33
tools are encourage this so for example
2:03:33
tools are encourage this so for example a common
2:03:35
a common
2:03:35
a common Tableau bi tool encourage users of
2:03:39
Tableau bi tool encourage users of
2:03:39
Tableau bi tool encourage users of extracts data science users are used to
2:03:43
extracts data science users are used to
2:03:43
extracts data science users are used to find this library or numpy
2:03:47
find this library or numpy
2:03:47
find this library or numpy this is their statistical API so as a
2:03:50
this is their statistical API so as a
2:03:50
this is their statistical API so as a general rule this common practice for
2:03:52
general rule this common practice for
2:03:52
general rule this common practice for data science users to perform their data
2:03:56
data science users to perform their data
2:03:56
data science users to perform their data science tasks on
2:03:58
science tasks on
2:03:58
science tasks on the CSV file or something like that so
2:04:01
the CSV file or something like that so
2:04:01
the CSV file or something like that so what suck we're getting as a result is
2:04:05
what suck we're getting as a result is
2:04:05
what suck we're getting as a result is user don't care about slow queries and
2:04:08
user don't care about slow queries and
2:04:08
user don't care about slow queries and why they if they run it once a day and
2:04:11
why they if they run it once a day and
2:04:11
why they if they run it once a day and they write an efficient scale that harms
2:04:13
they write an efficient scale that harms
2:04:13
they write an efficient scale that harms them and haram's other
2:04:16
them and haram's other
2:04:16
them and haram's other users of the system what I can do with
2:04:18
users of the system what I can do with
2:04:18
users of the system what I can do with that is education users tell them how to
2:04:22
that is education users tell them how to
2:04:22
that is education users tell them how to write efficient skill especially in
2:04:24
write efficient skill especially in
2:04:24
write efficient skill especially in cytos impose some restrictions there are
2:04:29
cytos impose some restrictions there are
2:04:29
cytos impose some restrictions there are statement demos obviously and there is
2:04:32
statement demos obviously and there is
2:04:32
statement demos obviously and there is also neat extension to postgres to
2:04:34
also neat extension to postgres to
2:04:34
also neat extension to postgres to prevent users to override this statement
2:04:36
prevent users to override this statement
2:04:36
prevent users to override this statement demos
2:04:37
demos
2:04:37
demos but you can also get go even further and
2:04:41
but you can also get go even further and
2:04:41
but you can also get go even further and Implement some sophisticated procedure
2:04:44
Implement some sophisticated procedure
2:04:44
Implement some sophisticated procedure to kill such problematic queries for
2:04:48
to kill such problematic queries for
2:04:48
to kill such problematic queries for those who try to work around those four
2:04:49
those who try to work around those four
2:04:49
those who try to work around those four crowns and you should obviously monitor
2:04:52
crowns and you should obviously monitor
2:04:52
crowns and you should obviously monitor how users use your analytical Warehouse
2:05:01
from our experiences experience there
2:05:04
from our experiences experience there
2:05:04
from our experiences experience there are two primary metrics to monitor first
2:05:08
are two primary metrics to monitor first
2:05:08
are two primary metrics to monitor first one is the memory consumption of course
2:05:12
uh you here you can see the our graphene
2:05:17
uh you here you can see the our graphene
2:05:17
uh you here you can see the our graphene dashboard of memory consumption peer pit
2:05:21
dashboard of memory consumption peer pit
2:05:21
dashboard of memory consumption peer pit and pair query within the deed and we
2:05:25
and pair query within the deed and we
2:05:25
and pair query within the deed and we collected usage using just previous
2:05:27
collected usage using just previous
2:05:27
collected usage using just previous adaptivity and PS2 nothing unusual here
2:05:32
adaptivity and PS2 nothing unusual here
2:05:32
adaptivity and PS2 nothing unusual here uh it is a little bit update outdated
2:05:35
uh it is a little bit update outdated
2:05:35
uh it is a little bit update outdated because it does not use Global pits yet
2:05:37
because it does not use Global pits yet
2:05:37
because it does not use Global pits yet but we are working on it and query
2:05:40
but we are working on it and query
2:05:40
but we are working on it and query execution times uh you already seen that
2:05:43
execution times uh you already seen that
2:05:43
execution times uh you already seen that nice
2:05:45
nice
2:05:45
nice graphic uh it is based on Pakistan's
2:05:49
graphic uh it is based on Pakistan's
2:05:49
graphic uh it is based on Pakistan's statement extension but we also collect
2:05:52
statement extension but we also collect
2:05:52
statement extension but we also collect slow queries directly from server locks
2:05:55
slow queries directly from server locks
2:05:55
slow queries directly from server locks via copy statement it sometimes it is a
2:05:59
via copy statement it sometimes it is a
2:05:59
via copy statement it sometimes it is a killer feature
2:06:01
killer feature
2:06:01
killer feature now what optimization do we have for uh
2:06:05
now what optimization do we have for uh
2:06:05
now what optimization do we have for uh users to suggest for Setters users
2:06:09
users to suggest for Setters users
2:06:09
users to suggest for Setters users should look out for distributed
2:06:11
should look out for distributed
2:06:11
should look out for distributed supplants and their queries uh it
2:06:13
supplants and their queries uh it
2:06:13
supplants and their queries uh it happens when you try to join distributed
2:06:15
happens when you try to join distributed
2:06:15
happens when you try to join distributed tables without usage of distribution
2:06:18
tables without usage of distribution
2:06:18
tables without usage of distribution columns using City for example uh it
2:06:21
columns using City for example uh it
2:06:21
columns using City for example uh it will lead to distributed Sublime in the
2:06:24
will lead to distributed Sublime in the
2:06:24
will lead to distributed Sublime in the query plan and essentially what it means
2:06:27
query plan and essentially what it means
2:06:27
query plan and essentially what it means is that status will have to pull all the
2:06:29
is that status will have to pull all the
2:06:29
is that status will have to pull all the rows from City to coordinator and then
2:06:32
rows from City to coordinator and then
2:06:32
rows from City to coordinator and then redistributing them back to workers
2:06:36
redistributing them back to workers
2:06:36
redistributing them back to workers uh by default uh this is called
2:06:38
uh by default uh this is called
2:06:38
uh by default uh this is called intermediate result and by default it is
2:06:40
intermediate result and by default it is
2:06:40
intermediate result and by default it is one gigabyte in size but as you can
2:06:42
one gigabyte in size but as you can
2:06:42
one gigabyte in size but as you can guess it's still haram's uh tremendously
2:06:44
guess it's still haram's uh tremendously
2:06:44
guess it's still haram's uh tremendously harms the performance tremendously so if
2:06:47
harms the performance tremendously so if
2:06:47
harms the performance tremendously so if you see this in your career plans you
2:06:50
you see this in your career plans you
2:06:50
you see this in your career plans you will need to either rewrite the query or
2:06:52
will need to either rewrite the query or
2:06:52
will need to either rewrite the query or uh redistribute your data
2:06:55
uh redistribute your data
2:06:55
uh redistribute your data now I haven't mentioned that but our
2:06:58
now I haven't mentioned that but our
2:06:58
now I haven't mentioned that but our data is partitioned and it is also a
2:07:01
data is partitioned and it is also a
2:07:01
data is partitioned and it is also a source of Errors because uh we found out
2:07:04
source of Errors because uh we found out
2:07:04
source of Errors because uh we found out that if you apply some transformation to
2:07:07
that if you apply some transformation to
2:07:07
that if you apply some transformation to partition column uh the podcast will not
2:07:10
partition column uh the podcast will not
2:07:10
partition column uh the podcast will not use the partition pronoun and your
2:07:13
use the partition pronoun and your
2:07:13
use the partition pronoun and your queries your query performance will
2:07:15
queries your query performance will
2:07:15
queries your query performance will suffer so just leave the partition
2:07:18
suffer so just leave the partition
2:07:18
suffer so just leave the partition column alone and indexes well default
2:07:21
column alone and indexes well default
2:07:21
column alone and indexes well default advice if you have some slow queries
2:07:24
advice if you have some slow queries
2:07:24
advice if you have some slow queries just create indexes if uh if they can be
2:07:28
just create indexes if uh if they can be
2:07:28
just create indexes if uh if they can be useful so okay now to conclusion uh here
2:07:33
useful so okay now to conclusion uh here
2:07:33
useful so okay now to conclusion uh here is the top wanted features to make our
2:07:35
is the top wanted features to make our
2:07:35
is the top wanted features to make our solution near ideal uh obviously it will
2:07:39
solution near ideal uh obviously it will
2:07:39
solution near ideal uh obviously it will be just on this Statistics obviously I
2:07:41
be just on this Statistics obviously I
2:07:41
be just on this Statistics obviously I would like to have multi-level schemas
2:07:43
would like to have multi-level schemas
2:07:43
would like to have multi-level schemas instead of databases uh so close
2:07:46
instead of databases uh so close
2:07:46
instead of databases uh so close database reference will be resolved
2:07:49
database reference will be resolved
2:07:49
database reference will be resolved essentially
2:07:50
essentially
2:07:50
essentially from cytus I would like to have sub
2:07:52
from cytus I would like to have sub
2:07:52
from cytus I would like to have sub partitions because currently it is not
2:07:54
partitions because currently it is not
2:07:54
partitions because currently it is not possible to create a partition of
2:07:55
possible to create a partition of
2:07:55
possible to create a partition of partition unfortunately
2:07:57
partition unfortunately
2:07:57
partition unfortunately and distributed mod views I assume I
2:08:01
and distributed mod views I assume I
2:08:01
and distributed mod views I assume I already
2:08:01
already
2:08:02
already described in colors why I would like to
2:08:05
described in colors why I would like to
2:08:05
described in colors why I would like to have that
2:08:06
have that
2:08:06
have that and as final thoughts cytus is a great
2:08:09
and as final thoughts cytus is a great
2:08:09
and as final thoughts cytus is a great partition a great extension for postgres
2:08:11
partition a great extension for postgres
2:08:11
partition a great extension for postgres it delivers exactly what is Promised in
2:08:14
it delivers exactly what is Promised in
2:08:14
it delivers exactly what is Promised in dogs and learning site but it is
2:08:16
dogs and learning site but it is
2:08:16
dogs and learning site but it is extension so you should be prepared that
2:08:20
extension so you should be prepared that
2:08:20
extension so you should be prepared that there are some edge cases uncovered
2:08:22
there are some edge cases uncovered
2:08:22
there are some edge cases uncovered there are some integration issues ETC so
2:08:25
there are some integration issues ETC so
2:08:25
there are some integration issues ETC so every every common issue for extension
2:08:29
every every common issue for extension
2:08:29
every every common issue for extension not a core product
2:08:30
not a core product
2:08:30
not a core product so you can find this presentation here
2:08:34
so you can find this presentation here
2:08:34
so you can find this presentation here on speaker decks speaker deck sorry and
2:08:38
on speaker decks speaker deck sorry and
2:08:38
on speaker decks speaker deck sorry and I am ready for your questions here and
2:08:41
I am ready for your questions here and
2:08:41
I am ready for your questions here and on Discord after the talk
2:08:43
on Discord after the talk
2:08:43
on Discord after the talk hurry hurry that was pretty uh
2:08:47
hurry hurry that was pretty uh
2:08:47
hurry hurry that was pretty uh interesting a flashback to serve many
2:08:48
interesting a flashback to serve many
2:08:48
interesting a flashback to serve many times I've written things like
2:08:50
times I've written things like
2:08:50
times I've written things like monitoring and custom implementations
2:08:52
monitoring and custom implementations
2:08:52
monitoring and custom implementations like that it's really good
2:08:56
I was really excited when you said um
2:09:00
I was really excited when you said um
2:09:00
I was really excited when you said um when you accepted the talk when you
2:09:02
when you accepted the talk when you
2:09:02
when you accepted the talk when you submitted the proposal when you said you
2:09:04
submitted the proposal when you said you
2:09:04
submitted the proposal when you said you were able to give a live stream version
2:09:06
were able to give a live stream version
2:09:06
were able to give a live stream version of the talk not just an on-demand talk
2:09:07
of the talk not just an on-demand talk
2:09:07
of the talk not just an on-demand talk because it's always fun for me to hear
2:09:10
because it's always fun for me to hear
2:09:10
because it's always fun for me to hear how you know people are using cytus in
2:09:12
how you know people are using cytus in
2:09:12
how you know people are using cytus in the real world in the wild
2:09:15
the real world in the wild
2:09:15
the real world in the wild oh yeah it was my pleasure because I
2:09:17
oh yeah it was my pleasure because I
2:09:17
oh yeah it was my pleasure because I wanted to let it out from my chest so so
2:09:20
wanted to let it out from my chest so so
2:09:20
wanted to let it out from my chest so so many years
2:09:22
many years
2:09:22
many years this is just a tip of the book actually
2:09:24
this is just a tip of the book actually
2:09:24
this is just a tip of the book actually because there are so much more yeah I
2:09:27
because there are so much more yeah I
2:09:27
because there are so much more yeah I was gonna say so I noticed on your
2:09:28
was gonna say so I noticed on your
2:09:28
was gonna say so I noticed on your slides that you said 2017 like that I
2:09:30
slides that you said 2017 like that I
2:09:30
slides that you said 2017 like that I mean that's the before times of so many
2:09:32
mean that's the before times of so many
2:09:32
mean that's the before times of so many things uh so what was the first version
2:09:34
things uh so what was the first version
2:09:34
things uh so what was the first version that you deployed this on with cytus uh
2:09:38
that you deployed this on with cytus uh
2:09:38
that you deployed this on with cytus uh I assume uh 7.0 7. okay so that's way
2:09:42
I assume uh 7.0 7. okay so that's way
2:09:42
I assume uh 7.0 7. okay so that's way back side is 7.0 yeah that's yes
2:09:45
back side is 7.0 yeah that's yes
2:09:45
back side is 7.0 yeah that's yes 7.0 so that's that's I mean that's way
2:09:48
7.0 so that's that's I mean that's way
2:09:48
7.0 so that's that's I mean that's way back uh you must have been ecstatic when
2:09:51
back uh you must have been ecstatic when
2:09:51
back uh you must have been ecstatic when Microsoft like fully open sourced cytus
2:09:54
Microsoft like fully open sourced cytus
2:09:54
Microsoft like fully open sourced cytus you know after the the purchase and all
2:09:56
you know after the the purchase and all
2:09:56
you know after the the purchase and all that uh you have no idea
2:09:59
that uh you have no idea
2:09:59
that uh you have no idea that was for us it was like our local
2:10:03
that was for us it was like our local
2:10:03
that was for us it was like our local celebration yeah and I imagine uh
2:10:05
celebration yeah and I imagine uh
2:10:05
celebration yeah and I imagine uh because I've run local scientists uh
2:10:07
because I've run local scientists uh
2:10:07
because I've run local scientists uh before for folks and uh The Shard
2:10:09
before for folks and uh The Shard
2:10:09
before for folks and uh The Shard rebalancer to me was also like one of
2:10:11
rebalancer to me was also like one of
2:10:11
rebalancer to me was also like one of the key pieces that like once that was
2:10:13
the key pieces that like once that was
2:10:13
the key pieces that like once that was put out there like it was kind of a game
2:10:15
put out there like it was kind of a game
2:10:15
put out there like it was kind of a game changer for people who wanted to do
2:10:16
changer for people who wanted to do
2:10:16
changer for people who wanted to do silos well once again once again as I
2:10:19
silos well once again once again as I
2:10:19
silos well once again once again as I said in the presentation it's not uh
2:10:21
said in the presentation it's not uh
2:10:21
said in the presentation it's not uh Silver Bullet it has its own issues but
2:10:23
Silver Bullet it has its own issues but
2:10:23
Silver Bullet it has its own issues but yeah still it is so much easier now
2:10:28
yeah still it is so much easier now
2:10:28
yeah still it is so much easier now so um for people who are just joining I
2:10:31
so um for people who are just joining I
2:10:31
so um for people who are just joining I just want to clarify cytus the extension
2:10:33
just want to clarify cytus the extension
2:10:33
just want to clarify cytus the extension did go open source back in 2016. when it
2:10:36
did go open source back in 2016. when it
2:10:36
did go open source back in 2016. when it was first unforked right and became a
2:10:38
was first unforked right and became a
2:10:38
was first unforked right and became a postgres extension but there were a few
2:10:40
postgres extension but there were a few
2:10:40
postgres extension but there were a few Enterprise features that were not open
2:10:42
Enterprise features that were not open
2:10:42
Enterprise features that were not open source at that point so the the going
2:10:44
source at that point so the the going
2:10:44
source at that point so the the going fully open source that was oh gosh 2021.
2:10:51
I remember that day there you go put it
2:10:54
I remember that day there you go put it
2:10:54
I remember that day there you go put it on the calendar all right well thank you
2:10:56
on the calendar all right well thank you
2:10:56
on the calendar all right well thank you so much uh I think there are some more
2:10:58
so much uh I think there are some more
2:10:58
so much uh I think there are some more questions in the Discord as to those
2:11:00
questions in the Discord as to those
2:11:00
questions in the Discord as to those things tend to do uh but once again
2:11:03
things tend to do uh but once again
2:11:03
things tend to do uh but once again thanks for speaking and uh it was
2:11:05
thanks for speaking and uh it was
2:11:05
thanks for speaking and uh it was awesome
2:11:08
thank you thank you so much Ivan ciao
2:11:11
thank you thank you so much Ivan ciao
2:11:11
thank you thank you so much Ivan ciao we'll see you on the Discord yeah
2:11:12
we'll see you on the Discord yeah
2:11:12
we'll see you on the Discord yeah obviously but
2:11:15
obviously but
2:11:15
obviously but all right uh and who do we have up next
2:11:19
all right uh and who do we have up next
2:11:19
all right uh and who do we have up next Pamela Fox welcome
2:11:22
Pamela Fox welcome
2:11:22
Pamela Fox welcome hello
2:11:26
hi Pamela is a cloud advocate in the
2:11:30
hi Pamela is a cloud advocate in the
2:11:30
hi Pamela is a cloud advocate in the python team at Microsoft and uh you're
2:11:33
python team at Microsoft and uh you're
2:11:33
python team at Microsoft and uh you're here to give us a talk today about using
2:11:35
here to give us a talk today about using
2:11:35
here to give us a talk today about using postgres on Azure using bicep
2:11:39
postgres on Azure using bicep
2:11:39
postgres on Azure using bicep um so for those of us who are not at
2:11:41
um so for those of us who are not at
2:11:41
um so for those of us who are not at Microsoft uh the title Cloud Advocate
2:11:44
Microsoft uh the title Cloud Advocate
2:11:44
Microsoft uh the title Cloud Advocate sounds a little ephemeral or something
2:11:45
sounds a little ephemeral or something
2:11:45
sounds a little ephemeral or something uh can you explain what that actually
2:11:47
uh can you explain what that actually
2:11:47
uh can you explain what that actually does what do you actually do yeah so my
2:11:50
does what do you actually do yeah so my
2:11:50
does what do you actually do yeah so my my job is like to Advocate on behalf of
2:11:53
my job is like to Advocate on behalf of
2:11:53
my job is like to Advocate on behalf of python Developers for the cloud you
2:11:56
python Developers for the cloud you
2:11:56
python Developers for the cloud you could also think of it as advocating on
2:11:57
could also think of it as advocating on
2:11:57
could also think of it as advocating on behalf of Microsoft I guess we kind of
2:11:59
behalf of Microsoft I guess we kind of
2:11:59
behalf of Microsoft I guess we kind of advocate on both fronts but I think it
2:12:01
advocate on both fronts but I think it
2:12:01
advocate on both fronts but I think it is trying to help python developers be
2:12:03
is trying to help python developers be
2:12:03
is trying to help python developers be successful with azir so whether that's
2:12:05
successful with azir so whether that's
2:12:05
successful with azir so whether that's you know using the Azure SDK or hosting
2:12:07
you know using the Azure SDK or hosting
2:12:07
you know using the Azure SDK or hosting things on Azure you know other things
2:12:09
things on Azure you know other things
2:12:09
things on Azure you know other things you can do with this year so trying to
2:12:11
you can do with this year so trying to
2:12:11
you can do with this year so trying to make it easier for python developers
2:12:12
make it easier for python developers
2:12:12
make it easier for python developers because I'm a python developer and I
2:12:14
because I'm a python developer and I
2:12:14
because I'm a python developer and I wanted to be better for them well and as
2:12:16
wanted to be better for them well and as
2:12:16
wanted to be better for them well and as I've been getting to know you Pamela it
2:12:18
I've been getting to know you Pamela it
2:12:18
I've been getting to know you Pamela it seems like you've been involved in
2:12:20
seems like you've been involved in
2:12:20
seems like you've been involved in helping people understand technology and
2:12:23
helping people understand technology and
2:12:23
helping people understand technology and understand how to solve problems for a
2:12:24
understand how to solve problems for a
2:12:24
understand how to solve problems for a while like you worked at Khan Academy
2:12:26
while like you worked at Khan Academy
2:12:26
while like you worked at Khan Academy didn't you
2:12:27
didn't you
2:12:27
didn't you yeah I created the computer programming
2:12:29
yeah I created the computer programming
2:12:29
yeah I created the computer programming curriculum for Khan Academy wow which it
2:12:31
curriculum for Khan Academy wow which it
2:12:31
curriculum for Khan Academy wow which it was everything but python
2:12:33
was everything but python
2:12:33
was everything but python JavaScript HTML CSS and SQL so I did
2:12:37
JavaScript HTML CSS and SQL so I did
2:12:37
JavaScript HTML CSS and SQL so I did that was um that was actually the first
2:12:39
that was um that was actually the first
2:12:39
that was um that was actually the first time I taught SQL was for Khan Academy
2:12:41
time I taught SQL was for Khan Academy
2:12:41
time I taught SQL was for Khan Academy and it was very fun to get into uh SQL
2:12:45
and it was very fun to get into uh SQL
2:12:45
and it was very fun to get into uh SQL awesome well without further Ado I think
2:12:49
awesome well without further Ado I think
2:12:49
awesome well without further Ado I think I think you should dive into your talk
2:12:50
I think you should dive into your talk
2:12:50
I think you should dive into your talk let's go
2:12:52
let's go
2:12:52
let's go all right
2:12:55
all right well hi everyone so since my
2:12:57
all right well hi everyone so since my
2:12:57
all right well hi everyone so since my job is to make azira better for Python
2:12:59
job is to make azira better for Python
2:12:59
job is to make azira better for Python devs and python devs love postgres I've
2:13:02
devs and python devs love postgres I've
2:13:02
devs and python devs love postgres I've been deploying hundreds of postgres
2:13:04
been deploying hundreds of postgres
2:13:04
been deploying hundreds of postgres servers uh you know for the last year so
2:13:07
servers uh you know for the last year so
2:13:07
servers uh you know for the last year so that I've been at Microsoft and I fell
2:13:09
that I've been at Microsoft and I fell
2:13:09
that I've been at Microsoft and I fell in love with using bicep to deploy them
2:13:10
in love with using bicep to deploy them
2:13:10
in love with using bicep to deploy them so I want to share that that love with
2:13:12
so I want to share that that love with
2:13:12
so I want to share that that love with all of you so my talk is for anyone who
2:13:15
all of you so my talk is for anyone who
2:13:15
all of you so my talk is for anyone who deploys postgres service to Azure or
2:13:18
deploys postgres service to Azure or
2:13:18
deploys postgres service to Azure or anyone who's just generally interested
2:13:19
anyone who's just generally interested
2:13:19
anyone who's just generally interested in using infrastructure as code as a
2:13:22
in using infrastructure as code as a
2:13:22
in using infrastructure as code as a deployment mechanism you know for any
2:13:24
deployment mechanism you know for any
2:13:24
deployment mechanism you know for any platform
2:13:26
so let's get into it so we're going to
2:13:29
so let's get into it so we're going to
2:13:29
so let's get into it so we're going to talk about using using bicep muscles to
2:13:34
talk about using using bicep muscles to
2:13:34
talk about using using bicep muscles to deploy postgres servers to azure
2:13:40
so just want to start off with
2:13:41
so just want to start off with
2:13:41
so just want to start off with clarification because there are multiple
2:13:43
clarification because there are multiple
2:13:43
clarification because there are multiple managed services for postgres on Azure
2:13:46
managed services for postgres on Azure
2:13:46
managed services for postgres on Azure uh there's the Azure database for
2:13:47
uh there's the Azure database for
2:13:47
uh there's the Azure database for postgres single server that was
2:13:49
postgres single server that was
2:13:49
postgres single server that was Microsoft's original offering it's no
2:13:51
Microsoft's original offering it's no
2:13:51
Microsoft's original offering it's no longer recommended for new apps so I
2:13:54
longer recommended for new apps so I
2:13:54
longer recommended for new apps so I basically just never never look at it
2:13:57
basically just never never look at it
2:13:57
basically just never never look at it um what they introduced since then is
2:13:59
um what they introduced since then is
2:13:59
um what they introduced since then is azure database for postgres flexible
2:14:01
azure database for postgres flexible
2:14:01
azure database for postgres flexible server
2:14:02
server
2:14:02
server and this one is another fully managed
2:14:04
and this one is another fully managed
2:14:04
and this one is another fully managed service but it offers vertical scaling
2:14:06
service but it offers vertical scaling
2:14:06
service but it offers vertical scaling and it also stays up to date with the
2:14:09
and it also stays up to date with the
2:14:09
and it also stays up to date with the latest postgres versions
2:14:10
latest postgres versions
2:14:10
latest postgres versions and there is also as your Cosmos to be
2:14:13
and there is also as your Cosmos to be
2:14:13
and there is also as your Cosmos to be for postgres which is cytus so it's uh
2:14:16
for postgres which is cytus so it's uh
2:14:16
for postgres which is cytus so it's uh you know a distributed database using
2:14:18
you know a distributed database using
2:14:18
you know a distributed database using postgres and the situs extension and
2:14:20
postgres and the situs extension and
2:14:20
postgres and the situs extension and that can scale horizontally as you may
2:14:23
that can scale horizontally as you may
2:14:23
that can scale horizontally as you may know if you're watching cytoscon
2:14:25
know if you're watching cytoscon
2:14:25
know if you're watching cytoscon um I'm going to be focusing today on the
2:14:27
um I'm going to be focusing today on the
2:14:27
um I'm going to be focusing today on the flexible server offering and um you know
2:14:30
flexible server offering and um you know
2:14:30
flexible server offering and um you know part of that is just due to what was
2:14:32
part of that is just due to what was
2:14:32
part of that is just due to what was available when I started Microsoft but
2:14:34
available when I started Microsoft but
2:14:34
available when I started Microsoft but it's also just a good fit for what
2:14:36
it's also just a good fit for what
2:14:36
it's also just a good fit for what python developers tend to use but
2:14:37
python developers tend to use but
2:14:37
python developers tend to use but everything I'm talking about today
2:14:38
everything I'm talking about today
2:14:38
everything I'm talking about today should be fairly applicable to the
2:14:42
should be fairly applicable to the
2:14:42
should be fairly applicable to the cosmos DB for postgres offering as well
2:14:44
cosmos DB for postgres offering as well
2:14:44
cosmos DB for postgres offering as well and if you're interested in trying to
2:14:46
and if you're interested in trying to
2:14:46
and if you're interested in trying to figure out the difference between them
2:14:47
figure out the difference between them
2:14:47
figure out the difference between them there's some you know links there that
2:14:49
there's some you know links there that
2:14:49
there's some you know links there that you can read more
2:14:51
you can read more
2:14:51
you can read more so there are many ways to display a
2:14:54
so there are many ways to display a
2:14:54
so there are many ways to display a postgres server to Azure and I have
2:14:57
postgres server to Azure and I have
2:14:57
postgres server to Azure and I have experimented with all those ways and
2:14:58
experimented with all those ways and
2:14:58
experimented with all those ways and that's how I came to realize which I
2:14:59
that's how I came to realize which I
2:15:00
that's how I came to realize which I like and don't like so uh you know let's
2:15:02
like and don't like so uh you know let's
2:15:02
like and don't like so uh you know let's let's look at those ways one way is to
2:15:05
let's look at those ways one way is to
2:15:05
let's look at those ways one way is to use the Azure portal and that's where
2:15:07
use the Azure portal and that's where
2:15:07
use the Azure portal and that's where you go to porto.azier.com you create
2:15:09
you go to porto.azier.com you create
2:15:09
you go to porto.azier.com you create your you know free as your account
2:15:11
your you know free as your account
2:15:11
your you know free as your account create a subscription and then you say
2:15:13
create a subscription and then you say
2:15:13
create a subscription and then you say okay I want to create a flexible server
2:15:15
okay I want to create a flexible server
2:15:15
okay I want to create a flexible server so then that's going to pop up this you
2:15:17
so then that's going to pop up this you
2:15:17
so then that's going to pop up this you know nice user interface with a form and
2:15:19
know nice user interface with a form and
2:15:19
know nice user interface with a form and one thing that I do really like about
2:15:21
one thing that I do really like about
2:15:21
one thing that I do really like about the portal is it that it does do this
2:15:22
the portal is it that it does do this
2:15:22
the portal is it that it does do this estimated cost that that is really cool
2:15:25
estimated cost that that is really cool
2:15:25
estimated cost that that is really cool I would love to have that from the
2:15:27
I would love to have that from the
2:15:27
I would love to have that from the command line and I've been asking the
2:15:28
command line and I've been asking the
2:15:28
command line and I've been asking the product teams about that
2:15:30
product teams about that
2:15:30
product teams about that um so that that to me is one of the best
2:15:32
um so that that to me is one of the best
2:15:32
um so that that to me is one of the best reasons for using the portal so the nice
2:15:34
reasons for using the portal so the nice
2:15:34
reasons for using the portal so the nice thing about using the portal is that you
2:15:35
thing about using the portal is that you
2:15:35
thing about using the portal is that you know you've got this uh you've got this
2:15:38
know you've got this uh you've got this
2:15:38
know you've got this uh you've got this UI here and it's really easy to get
2:15:40
UI here and it's really easy to get
2:15:40
UI here and it's really easy to get started and you just have to fill in
2:15:42
started and you just have to fill in
2:15:42
started and you just have to fill in some blanks and you get this cost
2:15:44
some blanks and you get this cost
2:15:44
some blanks and you get this cost estimate as well
2:15:46
estimate as well
2:15:46
estimate as well uh but it's also difficult to replicate
2:15:48
uh but it's also difficult to replicate
2:15:48
uh but it's also difficult to replicate so once I've made something in the
2:15:49
so once I've made something in the
2:15:49
so once I've made something in the portal and then I want to make another
2:15:51
portal and then I want to make another
2:15:51
portal and then I want to make another similar thing I have to remember like
2:15:53
similar thing I have to remember like
2:15:53
similar thing I have to remember like wait what did I you know what did I
2:15:55
wait what did I you know what did I
2:15:55
wait what did I you know what did I enter in that box I can't really
2:15:57
enter in that box I can't really
2:15:57
enter in that box I can't really remember what I put in there
2:15:59
remember what I put in there
2:15:59
remember what I put in there um so you know I don't tend to use it
2:16:01
um so you know I don't tend to use it
2:16:01
um so you know I don't tend to use it once I've once I've used it once I don't
2:16:03
once I've once I've used it once I don't
2:16:03
once I've once I've used it once I don't tend to use it again
2:16:05
tend to use it again
2:16:05
tend to use it again another option is using these CLI so
2:16:07
another option is using these CLI so
2:16:07
another option is using these CLI so there's the Azure CLI and then Azure
2:16:09
there's the Azure CLI and then Azure
2:16:09
there's the Azure CLI and then Azure Powershell so I have used the Azure CLI
2:16:12
Powershell so I have used the Azure CLI
2:16:12
Powershell so I have used the Azure CLI quite a bit
2:16:13
quite a bit
2:16:13
quite a bit um and uh you know it is nice because it
2:16:15
um and uh you know it is nice because it
2:16:15
um and uh you know it is nice because it can do you know pretty much everything
2:16:17
can do you know pretty much everything
2:16:17
can do you know pretty much everything you might want to do on Azure it can do
2:16:20
you might want to do on Azure it can do
2:16:20
you might want to do on Azure it can do uh so here we have a command to create a
2:16:22
uh so here we have a command to create a
2:16:22
uh so here we have a command to create a flexible server in a given location with
2:16:25
flexible server in a given location with
2:16:25
flexible server in a given location with a given name with this particular SKU
2:16:27
a given name with this particular SKU
2:16:27
a given name with this particular SKU and an admin user and password and uh
2:16:31
and an admin user and password and uh
2:16:31
and an admin user and password and uh and that will go ahead in and create
2:16:33
and that will go ahead in and create
2:16:33
and that will go ahead in and create that postgres server so you know the
2:16:35
that postgres server so you know the
2:16:35
that postgres server so you know the nice thing about command line options is
2:16:38
nice thing about command line options is
2:16:38
nice thing about command line options is that we can replicate them okay so I can
2:16:40
that we can replicate them okay so I can
2:16:40
that we can replicate them okay so I can say Hey you know colleague here's here's
2:16:43
say Hey you know colleague here's here's
2:16:43
say Hey you know colleague here's here's the command that I use to create the
2:16:45
the command that I use to create the
2:16:45
the command that I use to create the server if you want to create a similar
2:16:46
server if you want to create a similar
2:16:46
server if you want to create a similar server use this command
2:16:49
server use this command
2:16:49
server use this command however you know one thing that's not as
2:16:51
however you know one thing that's not as
2:16:51
however you know one thing that's not as nice is um that updating you know
2:16:55
nice is um that updating you know
2:16:55
nice is um that updating you know updating a server after the fact I
2:16:57
updating a server after the fact I
2:16:57
updating a server after the fact I actually have to use a different command
2:16:59
actually have to use a different command
2:16:59
actually have to use a different command for updating so now if I start to make
2:17:01
for updating so now if I start to make
2:17:01
for updating so now if I start to make updates now now I have to you know
2:17:03
updates now now I have to you know
2:17:03
updates now now I have to you know Cobble together this the sequence of
2:17:05
Cobble together this the sequence of
2:17:05
Cobble together this the sequence of command that I used and or put it
2:17:07
command that I used and or put it
2:17:07
command that I used and or put it together you know back into an original
2:17:09
together you know back into an original
2:17:09
together you know back into an original command so it is it's definitely you
2:17:12
command so it is it's definitely you
2:17:12
command so it is it's definitely you know much better than the portal in
2:17:13
know much better than the portal in
2:17:13
know much better than the portal in terms of repability uh but uh it still
2:17:16
terms of repability uh but uh it still
2:17:16
terms of repability uh but uh it still is difficult to maintain right so I'm
2:17:18
is difficult to maintain right so I'm
2:17:18
is difficult to maintain right so I'm looking to like what can I maintain what
2:17:20
looking to like what can I maintain what
2:17:20
looking to like what can I maintain what can I come back to six months from now
2:17:21
can I come back to six months from now
2:17:21
can I come back to six months from now and it'll just work
2:17:23
and it'll just work
2:17:23
and it'll just work so that brings me to arm templates and
2:17:25
so that brings me to arm templates and
2:17:25
so that brings me to arm templates and this is something that Azure has had for
2:17:27
this is something that Azure has had for
2:17:27
this is something that Azure has had for a while
2:17:28
a while
2:17:28
a while um and these are Json files that just
2:17:31
um and these are Json files that just
2:17:31
um and these are Json files that just declaratively describe the resources
2:17:33
declaratively describe the resources
2:17:33
declaratively describe the resources that we want to create this is just a
2:17:36
that we want to create this is just a
2:17:36
that we want to create this is just a snippet of the file and not the whole
2:17:38
snippet of the file and not the whole
2:17:38
snippet of the file and not the whole file uh but he's using Json and saying
2:17:41
file uh but he's using Json and saying
2:17:41
file uh but he's using Json and saying like Okay the type of resource we're
2:17:42
like Okay the type of resource we're
2:17:42
like Okay the type of resource we're making is a flexible server
2:17:44
making is a flexible server
2:17:44
making is a flexible server uh we're using this version of the API
2:17:47
uh we're using this version of the API
2:17:47
uh we're using this version of the API this is going to be the name of it this
2:17:49
this is going to be the name of it this
2:17:49
this is going to be the name of it this is the location this is the SKU and then
2:17:51
is the location this is the SKU and then
2:17:51
is the location this is the SKU and then we continue on with specifying the
2:17:53
we continue on with specifying the
2:17:53
we continue on with specifying the properties of that postgres server
2:17:56
properties of that postgres server
2:17:56
properties of that postgres server so the nice thing about this is that it
2:17:58
so the nice thing about this is that it
2:17:58
so the nice thing about this is that it you know everything is declared there
2:18:00
you know everything is declared there
2:18:00
you know everything is declared there and uh it is it is something we can
2:18:04
and uh it is it is something we can
2:18:04
and uh it is it is something we can repeat and we can share arm templates
2:18:06
repeat and we can share arm templates
2:18:06
repeat and we can share arm templates with you know whether other people so if
2:18:08
with you know whether other people so if
2:18:08
with you know whether other people so if you say oh you want to deploy the same
2:18:09
you say oh you want to deploy the same
2:18:09
you say oh you want to deploy the same server here you go and then if I come
2:18:11
server here you go and then if I come
2:18:12
server here you go and then if I come back to my project later I could just
2:18:14
back to my project later I could just
2:18:14
back to my project later I could just use that that arm template and uh and
2:18:16
use that that arm template and uh and
2:18:16
use that that arm template and uh and I'd be able to deploy it
2:18:18
I'd be able to deploy it
2:18:18
I'd be able to deploy it the annoying thing is it's Json you know
2:18:20
the annoying thing is it's Json you know
2:18:20
the annoying thing is it's Json you know I love Json as much as the next you know
2:18:24
I love Json as much as the next you know
2:18:24
I love Json as much as the next you know former JavaScript programmer but it is
2:18:27
former JavaScript programmer but it is
2:18:27
former JavaScript programmer but it is it is Jason and Jason is not a full
2:18:29
it is Jason and Jason is not a full
2:18:29
it is Jason and Jason is not a full programming language it's it's just a
2:18:31
programming language it's it's just a
2:18:31
programming language it's it's just a JavaScript object
2:18:34
JavaScript object
2:18:34
JavaScript object um and it can be hard to to actually use
2:18:37
um and it can be hard to to actually use
2:18:37
um and it can be hard to to actually use Json to to describe everything we might
2:18:40
Json to to describe everything we might
2:18:40
Json to to describe everything we might want to do and to to uh you know it's
2:18:44
want to do and to to uh you know it's
2:18:44
want to do and to to uh you know it's not a programming language so we can't
2:18:46
not a programming language so we can't
2:18:46
not a programming language so we can't conditionally create some properties
2:18:48
conditionally create some properties
2:18:48
conditionally create some properties based on some aspect of the environment
2:18:49
based on some aspect of the environment
2:18:49
based on some aspect of the environment right a very common thing is to have you
2:18:52
right a very common thing is to have you
2:18:52
right a very common thing is to have you know different staging environment than
2:18:54
know different staging environment than
2:18:54
know different staging environment than uh your production environment so your
2:18:55
uh your production environment so your
2:18:55
uh your production environment so your production environment might use a
2:18:57
production environment might use a
2:18:57
production environment might use a general purpose postgres server and this
2:18:59
general purpose postgres server and this
2:18:59
general purpose postgres server and this staging and Dev environment might use a
2:19:01
staging and Dev environment might use a
2:19:01
staging and Dev environment might use a burstable server which is much cheaper
2:19:05
burstable server which is much cheaper
2:19:05
burstable server which is much cheaper and you know based on those parameters
2:19:07
and you know based on those parameters
2:19:07
and you know based on those parameters you know you want to be able to say like
2:19:08
you know you want to be able to say like
2:19:08
you know you want to be able to say like okay we're going to deploy this mostly
2:19:11
okay we're going to deploy this mostly
2:19:11
okay we're going to deploy this mostly the same except except change a few
2:19:12
the same except except change a few
2:19:12
the same except except change a few things based on the environment and that
2:19:15
things based on the environment and that
2:19:15
things based on the environment and that can be difficult to do in just a Json
2:19:17
can be difficult to do in just a Json
2:19:17
can be difficult to do in just a Json file
2:19:18
file
2:19:18
file so
2:19:19
so
2:19:19
so that brings us to
2:19:22
that brings us to
2:19:22
that brings us to bicep so bicep is uh you know is the the
2:19:26
bicep so bicep is uh you know is the the
2:19:26
bicep so bicep is uh you know is the the solution to this
2:19:28
solution to this
2:19:28
solution to this and bicep is a uh it's a declarative
2:19:32
and bicep is a uh it's a declarative
2:19:32
and bicep is a uh it's a declarative language or it's also you know we call
2:19:34
language or it's also you know we call
2:19:34
language or it's also you know we call it a domain specific language it's
2:19:35
it a domain specific language it's
2:19:35
it a domain specific language it's specifically for deploying resources to
2:19:39
specifically for deploying resources to
2:19:39
specifically for deploying resources to azure uh you can also call it an
2:19:42
azure uh you can also call it an
2:19:42
azure uh you can also call it an infrastructure as code language so the
2:19:44
infrastructure as code language so the
2:19:44
infrastructure as code language so the idea is that you're describing your
2:19:46
idea is that you're describing your
2:19:46
idea is that you're describing your infrastructure as code it's very similar
2:19:48
infrastructure as code it's very similar
2:19:48
infrastructure as code it's very similar to terraform and terraform's the one
2:19:50
to terraform and terraform's the one
2:19:50
to terraform and terraform's the one that you know most people are familiar
2:19:51
that you know most people are familiar
2:19:51
that you know most people are familiar with because uh you can use terraform
2:19:53
with because uh you can use terraform
2:19:53
with because uh you can use terraform across all Cloud providers you can use
2:19:55
across all Cloud providers you can use
2:19:55
across all Cloud providers you can use terraform on Azure as well uh so if you
2:19:58
terraform on Azure as well uh so if you
2:19:58
terraform on Azure as well uh so if you do prefer terraform you can you know
2:20:00
do prefer terraform you can you know
2:20:00
do prefer terraform you can you know probably use some of the learnings from
2:20:02
probably use some of the learnings from
2:20:02
probably use some of the learnings from this talk but use it for terraform
2:20:03
this talk but use it for terraform
2:20:03
this talk but use it for terraform instead I do prefer to use bicep because
2:20:06
instead I do prefer to use bicep because
2:20:06
instead I do prefer to use bicep because it is it is targeted for Azure so it
2:20:09
it is it is targeted for Azure so it
2:20:09
it is it is targeted for Azure so it just tends to be a bit easier and it's
2:20:12
just tends to be a bit easier and it's
2:20:12
just tends to be a bit easier and it's and we know it's going to be up to date
2:20:14
and we know it's going to be up to date
2:20:14
and we know it's going to be up to date with Azure and it's got a lot of tolling
2:20:16
with Azure and it's got a lot of tolling
2:20:16
with Azure and it's got a lot of tolling that really helps you when you're an
2:20:18
that really helps you when you're an
2:20:18
that really helps you when you're an Azure developer uh but terraform is also
2:20:21
Azure developer uh but terraform is also
2:20:21
Azure developer uh but terraform is also a possibility and would have a lot of
2:20:23
a possibility and would have a lot of
2:20:23
a possibility and would have a lot of the benefits here so this is an
2:20:24
the benefits here so this is an
2:20:24
the benefits here so this is an infrastructure's code language we're
2:20:26
infrastructure's code language we're
2:20:26
infrastructure's code language we're going to declare
2:20:27
going to declare
2:20:27
going to declare uh what it is that we're creating and
2:20:30
uh what it is that we're creating and
2:20:30
uh what it is that we're creating and then you know deploy based on that uh so
2:20:32
then you know deploy based on that uh so
2:20:32
then you know deploy based on that uh so here you can say okay we're creating a
2:20:34
here you can say okay we're creating a
2:20:34
here you can say okay we're creating a server this is the name of server
2:20:35
server this is the name of server
2:20:35
server this is the name of server location SKU here are the properties uh
2:20:38
location SKU here are the properties uh
2:20:38
location SKU here are the properties uh so this is the entire bicep for the
2:20:40
so this is the entire bicep for the
2:20:40
so this is the entire bicep for the postgres server and this is kind of the
2:20:43
postgres server and this is kind of the
2:20:43
postgres server and this is kind of the minimal the minimal bicep that you would
2:20:45
minimal the minimal bicep that you would
2:20:45
minimal the minimal bicep that you would need to deploy a postgres server
2:20:48
need to deploy a postgres server
2:20:48
need to deploy a postgres server so how do we actually use this bicep oh
2:20:51
so how do we actually use this bicep oh
2:20:51
so how do we actually use this bicep oh well first let's talk about the fact
2:20:53
well first let's talk about the fact
2:20:53
well first let's talk about the fact that the bicep bicep is actually you
2:20:55
that the bicep bicep is actually you
2:20:55
that the bicep bicep is actually you know almost a full-featured language so
2:20:57
know almost a full-featured language so
2:20:57
know almost a full-featured language so uh it's you know it's not just like Json
2:21:00
uh it's you know it's not just like Json
2:21:00
uh it's you know it's not just like Json it also has parameters types uh logic
2:21:03
it also has parameters types uh logic
2:21:03
it also has parameters types uh logic Loops functions modules and that's where
2:21:06
Loops functions modules and that's where
2:21:06
Loops functions modules and that's where you get you know really nice ability to
2:21:08
you get you know really nice ability to
2:21:08
you get you know really nice ability to to refactor things and pass things into
2:21:10
to refactor things and pass things into
2:21:10
to refactor things and pass things into different modules uh so we're going to
2:21:13
different modules uh so we're going to
2:21:13
different modules uh so we're going to see examples of most of these uh most of
2:21:15
see examples of most of these uh most of
2:21:15
see examples of most of these uh most of these throughout the talk so you can see
2:21:17
these throughout the talk so you can see
2:21:17
these throughout the talk so you can see how you how it really helps the fact
2:21:19
how you how it really helps the fact
2:21:19
how you how it really helps the fact that bicep does have all these features
2:21:22
that bicep does have all these features
2:21:22
that bicep does have all these features of a language
2:21:26
all right so let's look at a
2:21:27
all right so let's look at a
2:21:27
all right so let's look at a parametrized bicep file so this is one
2:21:29
parametrized bicep file so this is one
2:21:29
parametrized bicep file so this is one of the you know features I was talking
2:21:30
of the you know features I was talking
2:21:30
of the you know features I was talking about parameters so this is really nice
2:21:33
about parameters so this is really nice
2:21:33
about parameters so this is really nice because I was saying you might have
2:21:34
because I was saying you might have
2:21:34
because I was saying you might have multiple environments you know a Dev
2:21:36
multiple environments you know a Dev
2:21:36
multiple environments you know a Dev environment a prod environment and you
2:21:38
environment a prod environment and you
2:21:38
environment a prod environment and you could use parameters for values that
2:21:40
could use parameters for values that
2:21:40
could use parameters for values that vary between those environments uh so
2:21:43
vary between those environments uh so
2:21:43
vary between those environments uh so here I have a parameter for the server
2:21:45
here I have a parameter for the server
2:21:45
here I have a parameter for the server name right because we're probably going
2:21:46
name right because we're probably going
2:21:46
name right because we're probably going to want a different server name for
2:21:48
to want a different server name for
2:21:48
to want a different server name for those two environments a location you
2:21:50
those two environments a location you
2:21:50
those two environments a location you might use different locations like you
2:21:52
might use different locations like you
2:21:52
might use different locations like you might you have a production location
2:21:53
might you have a production location
2:21:53
might you have a production location that's closer to your customers in a DAV
2:21:55
that's closer to your customers in a DAV
2:21:55
that's closer to your customers in a DAV environment that's closer to your devs
2:21:57
environment that's closer to your devs
2:21:57
environment that's closer to your devs uh and uh here I also have a param for
2:22:01
uh and uh here I also have a param for
2:22:01
uh and uh here I also have a param for the admin password and you notice also I
2:22:03
the admin password and you notice also I
2:22:03
the admin password and you notice also I use something called a decorator uh so
2:22:05
use something called a decorator uh so
2:22:05
use something called a decorator uh so there's these decorators that you can
2:22:06
there's these decorators that you can
2:22:06
there's these decorators that you can add to parameters in order to constrain
2:22:08
add to parameters in order to constrain
2:22:08
add to parameters in order to constrain them and so here I'm using the at secure
2:22:10
them and so here I'm using the at secure
2:22:10
them and so here I'm using the at secure decorator uh so that way bicep notes hey
2:22:12
decorator uh so that way bicep notes hey
2:22:12
decorator uh so that way bicep notes hey this is supposed to be a secure
2:22:13
this is supposed to be a secure
2:22:13
this is supposed to be a secure parameter so don't be you know like
2:22:15
parameter so don't be you know like
2:22:15
parameter so don't be you know like logging this and you know make sure that
2:22:18
logging this and you know make sure that
2:22:18
logging this and you know make sure that this is actually you know protected to
2:22:19
this is actually you know protected to
2:22:19
this is actually you know protected to some extent
2:22:20
some extent
2:22:20
some extent uh so you should definitely be using uh
2:22:23
uh so you should definitely be using uh
2:22:24
uh so you should definitely be using uh secure secure parameters uh for you know
2:22:26
secure secure parameters uh for you know
2:22:26
secure secure parameters uh for you know things like passwords
2:22:28
things like passwords
2:22:28
things like passwords um but you also should go way up and
2:22:30
um but you also should go way up and
2:22:30
um but you also should go way up and beyond for things like password so we'll
2:22:32
beyond for things like password so we'll
2:22:32
beyond for things like password so we'll talk more about that uh but uh yeah so
2:22:34
talk more about that uh but uh yeah so
2:22:34
talk more about that uh but uh yeah so there we go using params so this is a
2:22:36
there we go using params so this is a
2:22:36
there we go using params so this is a nice a nice starter file
2:22:39
nice a nice starter file
2:22:39
nice a nice starter file and now how how do I actually use this
2:22:41
and now how how do I actually use this
2:22:41
and now how how do I actually use this file so there is a command in the Azure
2:22:44
file so there is a command in the Azure
2:22:44
file so there is a command in the Azure CLI which is AZ deployment group create
2:22:48
CLI which is AZ deployment group create
2:22:48
CLI which is AZ deployment group create and I tell it what Resource Group is
2:22:51
and I tell it what Resource Group is
2:22:51
and I tell it what Resource Group is going to deploy the bicep resource to
2:22:53
going to deploy the bicep resource to
2:22:53
going to deploy the bicep resource to and I give it the template file so I say
2:22:55
and I give it the template file so I say
2:22:55
and I give it the template file so I say okay you're going to deploy to this
2:22:57
okay you're going to deploy to this
2:22:57
okay you're going to deploy to this Resource Group here's the file so I give
2:22:59
Resource Group here's the file so I give
2:22:59
Resource Group here's the file so I give it that bicep file that we just saw and
2:23:02
it that bicep file that we just saw and
2:23:02
it that bicep file that we just saw and it prompts for any parameters that
2:23:04
it prompts for any parameters that
2:23:04
it prompts for any parameters that haven't been specified so here since I
2:23:06
haven't been specified so here since I
2:23:06
haven't been specified so here since I didn't specify admin password it's
2:23:08
didn't specify admin password it's
2:23:08
didn't specify admin password it's prompting me for it so I'm just going to
2:23:10
prompting me for it so I'm just going to
2:23:10
prompting me for it so I'm just going to type it in you can also store passwords
2:23:12
type it in you can also store passwords
2:23:12
type it in you can also store passwords in a file uh you could fetch them from a
2:23:15
in a file uh you could fetch them from a
2:23:15
in a file uh you could fetch them from a key Vault so you have various options
2:23:17
key Vault so you have various options
2:23:17
key Vault so you have various options for how to pass parameters in but here
2:23:19
for how to pass parameters in but here
2:23:19
for how to pass parameters in but here I'm just going to type it in
2:23:21
I'm just going to type it in
2:23:21
I'm just going to type it in and not show you
2:23:23
and not show you
2:23:23
and not show you and then so it'll go off and it'll
2:23:26
and then so it'll go off and it'll
2:23:26
and then so it'll go off and it'll create that postgres server and it does
2:23:28
create that postgres server and it does
2:23:28
create that postgres server and it does take a few minutes because it's you know
2:23:29
take a few minutes because it's you know
2:23:29
take a few minutes because it's you know it's literally allocating some space for
2:23:31
it's literally allocating some space for
2:23:31
it's literally allocating some space for you in the cloud and carving that out
2:23:33
you in the cloud and carving that out
2:23:33
you in the cloud and carving that out for you
2:23:34
for you
2:23:34
for you um and postgres servers I do find take a
2:23:36
um and postgres servers I do find take a
2:23:36
um and postgres servers I do find take a little bit longer time to create than
2:23:38
little bit longer time to create than
2:23:38
little bit longer time to create than some of the other as your resources I
2:23:40
some of the other as your resources I
2:23:40
some of the other as your resources I create um so it could take a few minutes
2:23:42
create um so it could take a few minutes
2:23:42
create um so it could take a few minutes and then it comes back and what it comes
2:23:44
and then it comes back and what it comes
2:23:44
and then it comes back and what it comes back with is actually a bunch of jsons
2:23:46
back with is actually a bunch of jsons
2:23:46
back with is actually a bunch of jsons so we're back to Json again right it
2:23:47
so we're back to Json again right it
2:23:47
so we're back to Json again right it shows you a Json output of uh everything
2:23:50
shows you a Json output of uh everything
2:23:50
shows you a Json output of uh everything that got created and you can look
2:23:52
that got created and you can look
2:23:52
that got created and you can look through that output to you know see
2:23:54
through that output to you know see
2:23:54
through that output to you know see properties of it but typically what I do
2:23:56
properties of it but typically what I do
2:23:56
properties of it but typically what I do is I just go to the portal and I look at
2:23:58
is I just go to the portal and I look at
2:23:58
is I just go to the portal and I look at it
2:23:59
it
2:23:59
it uh you can also specify parameters on
2:24:01
uh you can also specify parameters on
2:24:01
uh you can also specify parameters on the command line that's another way to
2:24:02
the command line that's another way to
2:24:02
the command line that's another way to specifying parameters
2:24:05
specifying parameters
2:24:05
specifying parameters so once it deploys I like to go to the
2:24:07
so once it deploys I like to go to the
2:24:07
so once it deploys I like to go to the portal you know just check everything
2:24:09
portal you know just check everything
2:24:09
portal you know just check everything out make sure it is looking good so I
2:24:12
out make sure it is looking good so I
2:24:12
out make sure it is looking good so I can go in the portal and see okay yeah
2:24:14
can go in the portal and see okay yeah
2:24:14
can go in the portal and see okay yeah it's here this it's got the right server
2:24:16
it's here this it's got the right server
2:24:16
it's here this it's got the right server name it's you know it's got everything
2:24:17
name it's you know it's got everything
2:24:17
name it's you know it's got everything expected I mean it you know it it works
2:24:20
expected I mean it you know it it works
2:24:20
expected I mean it you know it it works so we don't really have to double check
2:24:22
so we don't really have to double check
2:24:22
so we don't really have to double check in the portal but it's nice to see that
2:24:24
in the portal but it's nice to see that
2:24:24
in the portal but it's nice to see that it's there and you can poke around at it
2:24:26
it's there and you can poke around at it
2:24:26
it's there and you can poke around at it uh from the portal
2:24:29
uh from the portal
2:24:29
uh from the portal yeah
2:24:30
yeah
2:24:30
yeah okay so now let's go more into what we
2:24:34
okay so now let's go more into what we
2:24:34
okay so now let's go more into what we can do with bicep so that we can
2:24:36
can do with bicep so that we can
2:24:36
can do with bicep so that we can customize This postgres Server more
2:24:37
customize This postgres Server more
2:24:37
customize This postgres Server more because that was basically the minimal
2:24:39
because that was basically the minimal
2:24:39
because that was basically the minimal postgres server
2:24:41
postgres server
2:24:41
postgres server um but there's a lot more that we might
2:24:42
um but there's a lot more that we might
2:24:42
um but there's a lot more that we might want to do with it uh so for the rest of
2:24:45
want to do with it uh so for the rest of
2:24:45
want to do with it uh so for the rest of what we want to do I'm going to use
2:24:46
what we want to do I'm going to use
2:24:46
what we want to do I'm going to use child resources so these are resources
2:24:49
child resources so these are resources
2:24:49
child resources so these are resources that only exist in the scope of a parent
2:24:52
that only exist in the scope of a parent
2:24:52
that only exist in the scope of a parent resource in a bicep file so for postgres
2:24:55
resource in a bicep file so for postgres
2:24:55
resource in a bicep file so for postgres we can make child resources for
2:24:58
we can make child resources for
2:24:58
we can make child resources for declaring the administrators
2:25:00
declaring the administrators
2:25:00
declaring the administrators configurations databases firewall rules
2:25:03
configurations databases firewall rules
2:25:03
configurations databases firewall rules and migrations
2:25:06
and migrations
2:25:06
and migrations so let's take a look at making a
2:25:07
so let's take a look at making a
2:25:07
so let's take a look at making a database so by default Azure postgres
2:25:10
database so by default Azure postgres
2:25:10
database so by default Azure postgres flexible servers they have a database
2:25:13
flexible servers they have a database
2:25:13
flexible servers they have a database called postgres and they have system
2:25:15
called postgres and they have system
2:25:15
called postgres and they have system databases named Azure maintenance and
2:25:17
databases named Azure maintenance and
2:25:17
databases named Azure maintenance and Azure CIS so every flexible server will
2:25:20
Azure CIS so every flexible server will
2:25:20
Azure CIS so every flexible server will have those databases but you may want to
2:25:23
have those databases but you may want to
2:25:23
have those databases but you may want to have a different database you might want
2:25:25
have a different database you might want
2:25:25
have a different database you might want to have a database with a different name
2:25:27
to have a database with a different name
2:25:27
to have a database with a different name so what you can do is create a child
2:25:31
so what you can do is create a child
2:25:31
so what you can do is create a child resource underneath you know inside your
2:25:34
resource underneath you know inside your
2:25:34
resource underneath you know inside your postgres server resource in the bicep
2:25:36
postgres server resource in the bicep
2:25:36
postgres server resource in the bicep file so you can see that you know here
2:25:38
file so you can see that you know here
2:25:38
file so you can see that you know here we start with the postgres server and
2:25:40
we start with the postgres server and
2:25:40
we start with the postgres server and then nested inside that we have a
2:25:42
then nested inside that we have a
2:25:42
then nested inside that we have a database resource and we give that a
2:25:44
database resource and we give that a
2:25:44
database resource and we give that a name web app and that's all we need to
2:25:46
name web app and that's all we need to
2:25:46
name web app and that's all we need to do for a database child resource so now
2:25:49
do for a database child resource so now
2:25:49
do for a database child resource so now when we deploy this it's actually going
2:25:51
when we deploy this it's actually going
2:25:51
when we deploy this it's actually going to create that database as well as the
2:25:53
to create that database as well as the
2:25:53
to create that database as well as the default ones
2:25:55
default ones
2:25:55
default ones we can get a little fancier if we want
2:25:58
we can get a little fancier if we want
2:25:58
we can get a little fancier if we want to have multiple databases uh so I'm
2:26:01
to have multiple databases uh so I'm
2:26:01
to have multiple databases uh so I'm going to make a param that's an array
2:26:03
going to make a param that's an array
2:26:03
going to make a param that's an array and the array is going to have the names
2:26:06
and the array is going to have the names
2:26:06
and the array is going to have the names for each of the databases so we have web
2:26:08
for each of the databases so we have web
2:26:08
for each of the databases so we have web app and analytics and then I can I can
2:26:12
app and analytics and then I can I can
2:26:12
app and analytics and then I can I can use a for Loop uh to to create multiple
2:26:17
use a for Loop uh to to create multiple
2:26:17
use a for Loop uh to to create multiple child resources and it has this syntax
2:26:19
child resources and it has this syntax
2:26:19
child resources and it has this syntax you see here where you say you know
2:26:21
you see here where you say you know
2:26:21
you see here where you say you know equals and then the brackets and the and
2:26:24
equals and then the brackets and the and
2:26:24
equals and then the brackets and the and the for Loop there and and that way I'm
2:26:27
the for Loop there and and that way I'm
2:26:27
the for Loop there and and that way I'm actually creating multiple child
2:26:28
actually creating multiple child
2:26:28
actually creating multiple child resources so this is a really common use
2:26:31
resources so this is a really common use
2:26:31
resources so this is a really common use of arrays and for Loops is to make
2:26:34
of arrays and for Loops is to make
2:26:34
of arrays and for Loops is to make multiple child resources and we'll see
2:26:36
multiple child resources and we'll see
2:26:36
multiple child resources and we'll see this come up again soon so hopefully
2:26:38
this come up again soon so hopefully
2:26:38
this come up again soon so hopefully you're starting to see some of the
2:26:39
you're starting to see some of the
2:26:39
you're starting to see some of the advantages of really having more
2:26:41
advantages of really having more
2:26:41
advantages of really having more language features inside bicep
2:26:45
language features inside bicep
2:26:45
language features inside bicep so once we've done that and deployed the
2:26:47
so once we've done that and deployed the
2:26:47
so once we've done that and deployed the you know deployed This postgres Server
2:26:49
you know deployed This postgres Server
2:26:49
you know deployed This postgres Server that has databases we could go to the
2:26:50
that has databases we could go to the
2:26:50
that has databases we could go to the portal we could click on databases and
2:26:53
portal we could click on databases and
2:26:53
portal we could click on databases and now we see that the additional two
2:26:55
now we see that the additional two
2:26:55
now we see that the additional two databases are in there analytics and web
2:26:57
databases are in there analytics and web
2:26:57
databases are in there analytics and web app and we also see we can connect to
2:26:59
app and we also see we can connect to
2:26:59
app and we also see we can connect to those databases that's something cool
2:27:00
those databases that's something cool
2:27:00
those databases that's something cool you can do is you can actually connect
2:27:02
you can do is you can actually connect
2:27:02
you can do is you can actually connect to the databases inside the portal and
2:27:04
to the databases inside the portal and
2:27:04
to the databases inside the portal and you know Tinker with them there
2:27:08
uh next we can make fire owls okay so by
2:27:11
uh next we can make fire owls okay so by
2:27:11
uh next we can make fire owls okay so by default you need to figure out when
2:27:12
default you need to figure out when
2:27:12
default you need to figure out when you're deploying a poster server what
2:27:14
you're deploying a poster server what
2:27:14
you're deploying a poster server what your networking story is going to be uh
2:27:16
your networking story is going to be uh
2:27:16
your networking story is going to be uh the default is that it's publicly
2:27:19
the default is that it's publicly
2:27:19
the default is that it's publicly accessible but it's not accessible to
2:27:21
accessible but it's not accessible to
2:27:21
accessible but it's not accessible to any particular IPS so it's not really
2:27:23
any particular IPS so it's not really
2:27:23
any particular IPS so it's not really publicly accessible so you need to
2:27:25
publicly accessible so you need to
2:27:25
publicly accessible so you need to explicitly tell it what IPS you're going
2:27:27
explicitly tell it what IPS you're going
2:27:27
explicitly tell it what IPS you're going to allow it to be accessible from or
2:27:30
to allow it to be accessible from or
2:27:30
to allow it to be accessible from or explicitly tell it not to be accessible
2:27:31
explicitly tell it not to be accessible
2:27:31
explicitly tell it not to be accessible to any IPS so in this case we're going
2:27:34
to any IPS so in this case we're going
2:27:34
to any IPS so in this case we're going to create a child resource
2:27:36
to create a child resource
2:27:36
to create a child resource um of the of firewall rules and tell it
2:27:39
um of the of firewall rules and tell it
2:27:39
um of the of firewall rules and tell it to allow as your internal IPS and the
2:27:42
to allow as your internal IPS and the
2:27:42
to allow as your internal IPS and the way we do that is say the start IP
2:27:44
way we do that is say the start IP
2:27:44
way we do that is say the start IP address is 0.0.00
2:27:46
address is 0.0.00
2:27:46
address is 0.0.00 might have too many zeros there and then
2:27:48
might have too many zeros there and then
2:27:48
might have too many zeros there and then the end ipaters is the same and this
2:27:51
the end ipaters is the same and this
2:27:51
the end ipaters is the same and this will allow any Azure resource to be able
2:27:54
will allow any Azure resource to be able
2:27:54
will allow any Azure resource to be able to contact that postgres server and you
2:27:56
to contact that postgres server and you
2:27:56
to contact that postgres server and you have to keep in mind that's going to
2:27:57
have to keep in mind that's going to
2:27:57
have to keep in mind that's going to mean any of yours or resources also
2:28:00
mean any of yours or resources also
2:28:00
mean any of yours or resources also anybody else's your resources so you
2:28:03
anybody else's your resources so you
2:28:03
anybody else's your resources so you know this is there's a bit of a security
2:28:05
know this is there's a bit of a security
2:28:05
know this is there's a bit of a security hole here if that if there's you know
2:28:07
hole here if that if there's you know
2:28:07
hole here if that if there's you know someone God forbid that's doing
2:28:09
someone God forbid that's doing
2:28:09
someone God forbid that's doing something nefarious on his ear and who
2:28:11
something nefarious on his ear and who
2:28:11
something nefarious on his ear and who manages to figure out your username
2:28:12
manages to figure out your username
2:28:12
manages to figure out your username password
2:28:13
password
2:28:13
password there they are going to be able to
2:28:15
there they are going to be able to
2:28:15
there they are going to be able to access it from there as you as your
2:28:17
access it from there as you as your
2:28:17
access it from there as you as your resource so we will talk about using a
2:28:19
resource so we will talk about using a
2:28:19
resource so we will talk about using a v-net which is a better approach
2:28:21
v-net which is a better approach
2:28:21
v-net which is a better approach uh but many you know this this is a
2:28:24
uh but many you know this this is a
2:28:24
uh but many you know this this is a decent approach to start off with and
2:28:27
decent approach to start off with and
2:28:27
decent approach to start off with and um we can also allow our individual IPS
2:28:29
um we can also allow our individual IPS
2:28:29
um we can also allow our individual IPS so if we wanted from our Dev machine to
2:28:31
so if we wanted from our Dev machine to
2:28:31
so if we wanted from our Dev machine to be able to access the server we can
2:28:33
be able to access the server we can
2:28:33
be able to access the server we can allow we can do a firewall rule just for
2:28:35
allow we can do a firewall rule just for
2:28:35
allow we can do a firewall rule just for that IP so here I've got an array you
2:28:38
that IP so here I've got an array you
2:28:38
that IP so here I've got an array you know with my IP and you know my my
2:28:40
know with my IP and you know my my
2:28:40
know with my IP and you know my my colleague's IP and I say okay let you
2:28:43
colleague's IP and I say okay let you
2:28:43
colleague's IP and I say okay let you know let all these IPS in here so this
2:28:45
know let all these IPS in here so this
2:28:45
know let all these IPS in here so this is a common setup to allow a zero IPS
2:28:47
is a common setup to allow a zero IPS
2:28:47
is a common setup to allow a zero IPS plus some individual IPS and you do
2:28:49
plus some individual IPS and you do
2:28:49
plus some individual IPS and you do really need to keep in mind the security
2:28:50
really need to keep in mind the security
2:28:50
really need to keep in mind the security implications of doing that
2:28:53
implications of doing that
2:28:53
implications of doing that uh and once we do that we can check the
2:28:55
uh and once we do that we can check the
2:28:55
uh and once we do that we can check the portal and what you'll see you know what
2:28:57
portal and what you'll see you know what
2:28:57
portal and what you'll see you know what it looks like in the networking tab is
2:28:58
it looks like in the networking tab is
2:28:58
it looks like in the networking tab is that it says the connectivity method is
2:29:00
that it says the connectivity method is
2:29:00
that it says the connectivity method is public access and that is going to allow
2:29:03
public access and that is going to allow
2:29:03
public access and that is going to allow Public Access from any Azure service
2:29:05
Public Access from any Azure service
2:29:05
Public Access from any Azure service within Azure to the server and then it's
2:29:07
within Azure to the server and then it's
2:29:07
within Azure to the server and then it's also going to allow those individual IPS
2:29:09
also going to allow those individual IPS
2:29:09
also going to allow those individual IPS so that's what it looks like after
2:29:11
so that's what it looks like after
2:29:11
so that's what it looks like after deploying that bicep file
2:29:15
all right so yeah let's talk about
2:29:17
all right so yeah let's talk about
2:29:17
all right so yeah let's talk about making it more secure by injecting the
2:29:21
making it more secure by injecting the
2:29:21
making it more secure by injecting the postgres server inside a virtual network
2:29:23
postgres server inside a virtual network
2:29:23
postgres server inside a virtual network uh you know with whatever it wants to
2:29:25
uh you know with whatever it wants to
2:29:25
uh you know with whatever it wants to communicate with it like so with like a
2:29:27
communicate with it like so with like a
2:29:27
communicate with it like so with like a python web app right and it's a python
2:29:28
python web app right and it's a python
2:29:28
python web app right and it's a python web app postcard server they can chill
2:29:30
web app postcard server they can chill
2:29:30
web app postcard server they can chill in the v-net so how do we do that in
2:29:32
in the v-net so how do we do that in
2:29:32
in the v-net so how do we do that in bicep this is where we're going to get a
2:29:34
bicep this is where we're going to get a
2:29:34
bicep this is where we're going to get a lot uh bigger with our bicep because we
2:29:37
lot uh bigger with our bicep because we
2:29:37
lot uh bigger with our bicep because we do need to create multiple resources so
2:29:40
do need to create multiple resources so
2:29:40
do need to create multiple resources so we're going to have to make a virtual
2:29:42
we're going to have to make a virtual
2:29:42
we're going to have to make a virtual Network so it is your virtual Network
2:29:44
Network so it is your virtual Network
2:29:44
Network so it is your virtual Network um and is your private DNS Zone and then
2:29:47
um and is your private DNS Zone and then
2:29:47
um and is your private DNS Zone and then also whatever resource actually needs to
2:29:49
also whatever resource actually needs to
2:29:49
also whatever resource actually needs to access the postgres server so in this
2:29:51
access the postgres server so in this
2:29:51
access the postgres server so in this example I'm doing an app service web app
2:29:53
example I'm doing an app service web app
2:29:53
example I'm doing an app service web app but you could have as you know as your
2:29:55
but you could have as you know as your
2:29:55
but you could have as you know as your container app and as your function a VM
2:29:58
container app and as your function a VM
2:29:58
container app and as your function a VM whatever it is that needs access is is
2:30:00
whatever it is that needs access is is
2:30:00
whatever it is that needs access is is going to want to live inside that Vina
2:30:02
going to want to live inside that Vina
2:30:02
going to want to live inside that Vina as well
2:30:05
so the first thing we do is create that
2:30:07
so the first thing we do is create that
2:30:07
so the first thing we do is create that and create that v-net with a we have to
2:30:09
and create that v-net with a we have to
2:30:09
and create that v-net with a we have to tell it you know it's possible private
2:30:11
tell it you know it's possible private
2:30:11
tell it you know it's possible private address space so I gave it a range there
2:30:13
address space so I gave it a range there
2:30:13
address space so I gave it a range there I'm using the cidr notation
2:30:17
I'm using the cidr notation
2:30:17
I'm using the cidr notation and then next I create subnets inside
2:30:19
and then next I create subnets inside
2:30:19
and then next I create subnets inside that v-net and I delegate ranges within
2:30:23
that v-net and I delegate ranges within
2:30:23
that v-net and I delegate ranges within that original address space I delegate
2:30:25
that original address space I delegate
2:30:26
that original address space I delegate ranges to particular types of azure
2:30:29
ranges to particular types of azure
2:30:29
ranges to particular types of azure services so here I'm saying that this
2:30:33
services so here I'm saying that this
2:30:33
services so here I'm saying that this range gets delegated to flexible servers
2:30:38
range gets delegated to flexible servers
2:30:38
range gets delegated to flexible servers uh so that's they're gonna get you know
2:30:40
uh so that's they're gonna get you know
2:30:40
uh so that's they're gonna get you know live in that part of the v-net
2:30:43
live in that part of the v-net
2:30:43
live in that part of the v-net and then I also need a subnet for the
2:30:46
and then I also need a subnet for the
2:30:46
and then I also need a subnet for the other resource so in that case it's I
2:30:47
other resource so in that case it's I
2:30:48
other resource so in that case it's I said server Farms which is actually app
2:30:49
said server Farms which is actually app
2:30:49
said server Farms which is actually app service web apps uh so I said okay this
2:30:52
service web apps uh so I said okay this
2:30:52
service web apps uh so I said okay this this other range uh non-overlapping
2:30:55
this other range uh non-overlapping
2:30:55
this other range uh non-overlapping range is going to be dedicated to these
2:30:57
range is going to be dedicated to these
2:30:57
range is going to be dedicated to these server Farms
2:31:00
and now uh in order to let the other
2:31:03
and now uh in order to let the other
2:31:03
and now uh in order to let the other servers so my app service web app is
2:31:05
servers so my app service web app is
2:31:05
servers so my app service web app is going to be publicly accessible so it
2:31:07
going to be publicly accessible so it
2:31:07
going to be publicly accessible so it doesn't need a private DNS right because
2:31:09
doesn't need a private DNS right because
2:31:09
doesn't need a private DNS right because we're letting it I'm going to let anyone
2:31:10
we're letting it I'm going to let anyone
2:31:10
we're letting it I'm going to let anyone hit that up right no problem but I don't
2:31:13
hit that up right no problem but I don't
2:31:13
hit that up right no problem but I don't want my postgres server to be public
2:31:14
want my postgres server to be public
2:31:14
want my postgres server to be public accessible but I do want to be able to
2:31:16
accessible but I do want to be able to
2:31:16
accessible but I do want to be able to access it via a URL instead of an IP so
2:31:19
access it via a URL instead of an IP so
2:31:19
access it via a URL instead of an IP so what I do is create a private DNS Zone
2:31:22
what I do is create a private DNS Zone
2:31:22
what I do is create a private DNS Zone and the name there's some naming rules
2:31:25
and the name there's some naming rules
2:31:25
and the name there's some naming rules here so the name has to end in
2:31:28
here so the name has to end in
2:31:28
here so the name has to end in postgres.database.zero.com so you see
2:31:30
postgres.database.zero.com so you see
2:31:30
postgres.database.zero.com so you see that on the second line but it can't be
2:31:32
that on the second line but it can't be
2:31:32
that on the second line but it can't be the exact same as the uh as a database
2:31:35
the exact same as the uh as a database
2:31:35
the exact same as the uh as a database so here I have dot private Dot
2:31:38
so here I have dot private Dot
2:31:38
so here I have dot private Dot progress.database.zero.com and that
2:31:41
progress.database.zero.com and that
2:31:41
progress.database.zero.com and that satisfies the naming rules for this
2:31:43
satisfies the naming rules for this
2:31:43
satisfies the naming rules for this private DNS Zone and there's more
2:31:44
private DNS Zone and there's more
2:31:44
private DNS Zone and there's more information in the links on this slide
2:31:46
information in the links on this slide
2:31:46
information in the links on this slide because it is a little tricky
2:31:49
because it is a little tricky
2:31:49
because it is a little tricky uh so I create this private DNS Zone and
2:31:51
uh so I create this private DNS Zone and
2:31:51
uh so I create this private DNS Zone and then I link it to the virtual Network
2:31:54
then I link it to the virtual Network
2:31:54
then I link it to the virtual Network and that is what's going to allow the
2:31:56
and that is what's going to allow the
2:31:56
and that is what's going to allow the web app to be able to use a the URL to
2:32:00
web app to be able to use a the URL to
2:32:00
web app to be able to use a the URL to actually interact with the server
2:32:02
actually interact with the server
2:32:02
actually interact with the server despite the fact that the postgres
2:32:03
despite the fact that the postgres
2:32:03
despite the fact that the postgres server isn't technically accessible at
2:32:06
server isn't technically accessible at
2:32:06
server isn't technically accessible at that URL uh Azure will be able to look
2:32:08
that URL uh Azure will be able to look
2:32:08
that URL uh Azure will be able to look up and realize that there's a private
2:32:10
up and realize that there's a private
2:32:10
up and realize that there's a private DNS that's doing a mapping to a v-net
2:32:12
DNS that's doing a mapping to a v-net
2:32:12
DNS that's doing a mapping to a v-net instead
2:32:15
now I need to configure both the
2:32:17
now I need to configure both the
2:32:17
now I need to configure both the resources so that they know about the
2:32:18
resources so that they know about the
2:32:18
resources so that they know about the v-net so they get injected into the
2:32:20
v-net so they get injected into the
2:32:20
v-net so they get injected into the v-net as it's called so I add on the
2:32:23
v-net as it's called so I add on the
2:32:23
v-net as it's called so I add on the postgres server I add a network property
2:32:26
postgres server I add a network property
2:32:26
postgres server I add a network property and I add there the delegated subnet
2:32:29
and I add there the delegated subnet
2:32:29
and I add there the delegated subnet resource ID so I get that from the
2:32:33
resource ID so I get that from the
2:32:33
resource ID so I get that from the virtual Network object you know resource
2:32:35
virtual Network object you know resource
2:32:35
virtual Network object you know resource that was made before and then I also
2:32:38
that was made before and then I also
2:32:38
that was made before and then I also have the private DNS Zone arm resource
2:32:40
have the private DNS Zone arm resource
2:32:40
have the private DNS Zone arm resource ID and so that's what's linking it to
2:32:43
ID and so that's what's linking it to
2:32:43
ID and so that's what's linking it to the private DNS zone so that postgres
2:32:45
the private DNS zone so that postgres
2:32:45
the private DNS zone so that postgres server has to get connected to both
2:32:47
server has to get connected to both
2:32:47
server has to get connected to both these things both the v-net and the
2:32:49
these things both the v-net and the
2:32:49
these things both the v-net and the private DNS
2:32:51
private DNS
2:32:51
private DNS but then the web app only needs to be
2:32:53
but then the web app only needs to be
2:32:53
but then the web app only needs to be connected to the v-net so it gets a it
2:32:56
connected to the v-net so it gets a it
2:32:56
connected to the v-net so it gets a it gets connected to this you know with the
2:32:58
gets connected to this you know with the
2:32:58
gets connected to this you know with the subnet resource ID as well
2:33:00
subnet resource ID as well
2:33:00
subnet resource ID as well uh in this case using a child resource
2:33:04
uh in this case using a child resource
2:33:04
uh in this case using a child resource all right so now we have we have all the
2:33:07
all right so now we have we have all the
2:33:07
all right so now we have we have all the pieces and there's a lot there's a lot
2:33:09
pieces and there's a lot there's a lot
2:33:09
pieces and there's a lot there's a lot of a lot of pieces here and uh you know
2:33:13
of a lot of pieces here and uh you know
2:33:13
of a lot of pieces here and uh you know I I'm providing all this code will be
2:33:15
I I'm providing all this code will be
2:33:15
I I'm providing all this code will be accessible afterwards so you're welcome
2:33:17
accessible afterwards so you're welcome
2:33:17
accessible afterwards so you're welcome to you know to just take it if this is
2:33:19
to you know to just take it if this is
2:33:19
to you know to just take it if this is the thing you're trying to do
2:33:21
the thing you're trying to do
2:33:21
the thing you're trying to do uh so all together you know end up with
2:33:23
uh so all together you know end up with
2:33:23
uh so all together you know end up with a fairly long file that has all these
2:33:25
a fairly long file that has all these
2:33:25
a fairly long file that has all these resources in it and uh and you'll be
2:33:28
resources in it and uh and you'll be
2:33:28
resources in it and uh and you'll be able to follow the the links in these
2:33:29
able to follow the the links in these
2:33:29
able to follow the the links in these slides when I share them at the end uh
2:33:32
slides when I share them at the end uh
2:33:32
slides when I share them at the end uh but all together you're gonna you know
2:33:33
but all together you're gonna you know
2:33:33
but all together you're gonna you know have the file with all those resources
2:33:35
have the file with all those resources
2:33:35
have the file with all those resources and also have the web app declared in
2:33:37
and also have the web app declared in
2:33:37
and also have the web app declared in there and uh and then you can you know
2:33:39
there and uh and then you can you know
2:33:39
there and uh and then you can you know we can deploy that using the same
2:33:41
we can deploy that using the same
2:33:41
we can deploy that using the same command as before
2:33:43
command as before
2:33:43
command as before so once we deploy it we can check the
2:33:46
so once we deploy it we can check the
2:33:46
so once we deploy it we can check the portal and now we see we've got four
2:33:48
portal and now we see we've got four
2:33:48
portal and now we see we've got four resources deployed right the postgres
2:33:50
resources deployed right the postgres
2:33:50
resources deployed right the postgres server oh which I left out of that
2:33:52
server oh which I left out of that
2:33:52
server oh which I left out of that screenshot but we'll have a postgres
2:33:54
screenshot but we'll have a postgres
2:33:54
screenshot but we'll have a postgres server we'll have the app service we'll
2:33:56
server we'll have the app service we'll
2:33:56
server we'll have the app service we'll have the virtual Network we'll have a
2:33:57
have the virtual Network we'll have a
2:33:57
have the virtual Network we'll have a private NS Zone we also have an app
2:33:59
private NS Zone we also have an app
2:33:59
private NS Zone we also have an app service plan for the app service so you
2:34:01
service plan for the app service so you
2:34:01
service plan for the app service so you end up having a bunch of resources
2:34:03
end up having a bunch of resources
2:34:03
end up having a bunch of resources um so a v-net is a best practice from a
2:34:06
um so a v-net is a best practice from a
2:34:06
um so a v-net is a best practice from a security perspective it is a little more
2:34:08
security perspective it is a little more
2:34:08
security perspective it is a little more complicated to set up luckily I've done
2:34:10
complicated to set up luckily I've done
2:34:10
complicated to set up luckily I've done the work for you I'm sorry so you can
2:34:13
the work for you I'm sorry so you can
2:34:13
the work for you I'm sorry so you can you know take that work
2:34:14
you know take that work
2:34:15
you know take that work um it also can cost a little bit more so
2:34:17
um it also can cost a little bit more so
2:34:17
um it also can cost a little bit more so you can look at the cost for private DNS
2:34:19
you can look at the cost for private DNS
2:34:19
you can look at the cost for private DNS Zone and virtual network uh just to see
2:34:21
Zone and virtual network uh just to see
2:34:21
Zone and virtual network uh just to see if it's reasonable for you to use that
2:34:24
if it's reasonable for you to use that
2:34:24
if it's reasonable for you to use that as a security mechanism but it is
2:34:26
as a security mechanism but it is
2:34:26
as a security mechanism but it is definitely uh recommended from a
2:34:29
definitely uh recommended from a
2:34:29
definitely uh recommended from a security perspective
2:34:32
okay so uh let's talk about tips and
2:34:36
okay so uh let's talk about tips and
2:34:36
okay so uh let's talk about tips and tricks for writing bicep I've been
2:34:39
tricks for writing bicep I've been
2:34:39
tricks for writing bicep I've been writing quite a lot of bicep over the
2:34:41
writing quite a lot of bicep over the
2:34:41
writing quite a lot of bicep over the last eight months and I guess I've
2:34:42
last eight months and I guess I've
2:34:42
last eight months and I guess I've gotten a reputation as the bicep person
2:34:44
gotten a reputation as the bicep person
2:34:44
gotten a reputation as the bicep person on my team or at least you know the one
2:34:46
on my team or at least you know the one
2:34:46
on my team or at least you know the one that loves bicep uh so uh yeah so let's
2:34:49
that loves bicep uh so uh yeah so let's
2:34:49
that loves bicep uh so uh yeah so let's talk about some tips here
2:34:51
talk about some tips here
2:34:51
talk about some tips here so with postgresql I you know I showed
2:34:53
so with postgresql I you know I showed
2:34:53
so with postgresql I you know I showed username password I showed v-net there
2:34:56
username password I showed v-net there
2:34:56
username password I showed v-net there are other things to consider for
2:34:58
are other things to consider for
2:34:58
are other things to consider for security
2:34:59
security
2:34:59
security um so one thing that's more serious
2:35:00
um so one thing that's more serious
2:35:00
um so one thing that's more serious approach than username password is to
2:35:02
approach than username password is to
2:35:02
approach than username password is to use managed identity uh and so manage
2:35:05
use managed identity uh and so manage
2:35:05
use managed identity uh and so manage identity is you like you give an
2:35:07
identity is you like you give an
2:35:07
identity is you like you give an identity to your Azure server and you
2:35:09
identity to your Azure server and you
2:35:09
identity to your Azure server and you say it's a system of science identity
2:35:11
say it's a system of science identity
2:35:11
say it's a system of science identity and then it's allowed to it kind of kind
2:35:14
and then it's allowed to it kind of kind
2:35:14
and then it's allowed to it kind of kind of like creates username passwords on on
2:35:16
of like creates username passwords on on
2:35:16
of like creates username passwords on on demand it generates these tokens
2:35:19
demand it generates these tokens
2:35:19
demand it generates these tokens um and uh uses that instead so it's nice
2:35:22
um and uh uses that instead so it's nice
2:35:22
um and uh uses that instead so it's nice because you don't have to do Secret
2:35:23
because you don't have to do Secret
2:35:23
because you don't have to do Secret rotation it's kind of doing that for you
2:35:25
rotation it's kind of doing that for you
2:35:25
rotation it's kind of doing that for you so manage identity is generally a best
2:35:28
so manage identity is generally a best
2:35:28
so manage identity is generally a best practice uh it can be a little trickier
2:35:30
practice uh it can be a little trickier
2:35:30
practice uh it can be a little trickier to do in bicep I did recently work on a
2:35:33
to do in bicep I did recently work on a
2:35:33
to do in bicep I did recently work on a sample to uh that uses both bicep and
2:35:36
sample to uh that uses both bicep and
2:35:36
sample to uh that uses both bicep and the Azure CLI in order to do a postgres
2:35:39
the Azure CLI in order to do a postgres
2:35:39
the Azure CLI in order to do a postgres with managed identity so if you're
2:35:40
with managed identity so if you're
2:35:40
with managed identity so if you're interested in that check out that sample
2:35:42
interested in that check out that sample
2:35:42
interested in that check out that sample the ultimate is manage identity inside
2:35:44
the ultimate is manage identity inside
2:35:44
the ultimate is manage identity inside of unit and I think my colleague Aaron
2:35:46
of unit and I think my colleague Aaron
2:35:46
of unit and I think my colleague Aaron is going to be working on that
2:35:48
is going to be working on that
2:35:48
is going to be working on that uh
2:35:50
uh
2:35:50
uh location location location so the
2:35:52
location location location so the
2:35:52
location location location so the biggest problem I've had with deploying
2:35:54
biggest problem I've had with deploying
2:35:54
biggest problem I've had with deploying postgres servers is location constraints
2:35:57
postgres servers is location constraints
2:35:57
postgres servers is location constraints and this is actually mostly because I'm
2:35:59
and this is actually mostly because I'm
2:35:59
and this is actually mostly because I'm a Microsoft employee so uh Microsoft
2:36:01
a Microsoft employee so uh Microsoft
2:36:01
a Microsoft employee so uh Microsoft employees do have different location
2:36:03
employees do have different location
2:36:03
employees do have different location constraints than public folks so
2:36:06
constraints than public folks so
2:36:06
constraints than public folks so hopefully you all won't run into this
2:36:07
hopefully you all won't run into this
2:36:07
hopefully you all won't run into this but as a Microsoft employee if you do
2:36:09
but as a Microsoft employee if you do
2:36:09
but as a Microsoft employee if you do run into an error it might be because
2:36:10
run into an error it might be because
2:36:10
run into an error it might be because you're not permitted to deploy postgres
2:36:13
you're not permitted to deploy postgres
2:36:13
you're not permitted to deploy postgres in that location and in that case you
2:36:15
in that location and in that case you
2:36:15
in that location and in that case you should change the location for example
2:36:17
should change the location for example
2:36:17
should change the location for example East us2 Microsoft employees can't
2:36:19
East us2 Microsoft employees can't
2:36:19
East us2 Microsoft employees can't provision in that for postgres
2:36:22
provision in that for postgres
2:36:22
provision in that for postgres um hopefully if you're not a Microsoft
2:36:24
um hopefully if you're not a Microsoft
2:36:24
um hopefully if you're not a Microsoft employee you're just not going to run
2:36:25
employee you're just not going to run
2:36:25
employee you're just not going to run into that
2:36:25
into that
2:36:25
into that um and uh you know the thing is we're
2:36:27
um and uh you know the thing is we're
2:36:27
um and uh you know the thing is we're you know keeping things more free for
2:36:29
you know keeping things more free for
2:36:29
you know keeping things more free for the uh you know the non-marksoft
2:36:31
the uh you know the non-marksoft
2:36:31
the uh you know the non-marksoft employees out there so it's better for
2:36:34
employees out there so it's better for
2:36:34
employees out there so it's better for you
2:36:34
you
2:36:34
you uh another good thing to keep in mind is
2:36:37
uh another good thing to keep in mind is
2:36:37
uh another good thing to keep in mind is that you know with bicep one of the cool
2:36:39
that you know with bicep one of the cool
2:36:39
that you know with bicep one of the cool things is you think with bicep like oh
2:36:41
things is you think with bicep like oh
2:36:41
things is you think with bicep like oh okay I need to change my server
2:36:42
okay I need to change my server
2:36:42
okay I need to change my server configuration I'm just going to change
2:36:43
configuration I'm just going to change
2:36:43
configuration I'm just going to change my bicep and you know redeploy it you
2:36:47
my bicep and you know redeploy it you
2:36:47
my bicep and you know redeploy it you can't always change everything postgres
2:36:49
can't always change everything postgres
2:36:49
can't always change everything postgres actually has a number of constraints in
2:36:50
actually has a number of constraints in
2:36:50
actually has a number of constraints in terms of what can't be changed uh so you
2:36:53
terms of what can't be changed uh so you
2:36:53
terms of what can't be changed uh so you can't change the admin username once
2:36:55
can't change the admin username once
2:36:55
can't change the admin username once it's set it's set you can change the
2:36:56
it's set it's set you can change the
2:36:56
it's set it's set you can change the password which is good but you can't
2:36:57
password which is good but you can't
2:36:57
password which is good but you can't change the admin username uh you can't
2:37:00
change the admin username uh you can't
2:37:00
change the admin username uh you can't change the postgresql version so if you
2:37:02
change the postgresql version so if you
2:37:02
change the postgresql version so if you did want to update to postgres 14 you
2:37:05
did want to update to postgres 14 you
2:37:05
did want to update to postgres 14 you would need to you know go through the
2:37:07
would need to you know go through the
2:37:07
would need to you know go through the process of doing a a database backup and
2:37:10
process of doing a a database backup and
2:37:10
process of doing a a database backup and a restore to a new server the nice thing
2:37:12
a restore to a new server the nice thing
2:37:12
a restore to a new server the nice thing is that you could you know just copy and
2:37:14
is that you could you know just copy and
2:37:14
is that you could you know just copy and paste your bicep and now you've got a
2:37:16
paste your bicep and now you've got a
2:37:16
paste your bicep and now you've got a very similar server you don't even copy
2:37:18
very similar server you don't even copy
2:37:18
very similar server you don't even copy paste just change the parameter uh you
2:37:20
paste just change the parameter uh you
2:37:20
paste just change the parameter uh you can't change the location you can't
2:37:22
can't change the location you can't
2:37:22
can't change the location you can't change the networking option so you
2:37:23
change the networking option so you
2:37:23
change the networking option so you should decide from the get-go is do you
2:37:25
should decide from the get-go is do you
2:37:25
should decide from the get-go is do you want to go v-not or do you want to go
2:37:27
want to go v-not or do you want to go
2:37:27
want to go v-not or do you want to go public
2:37:28
public
2:37:28
public um so yeah you know in theory with bicep
2:37:31
um so yeah you know in theory with bicep
2:37:31
um so yeah you know in theory with bicep with most things you can most resources
2:37:33
with most things you can most resources
2:37:33
with most things you can most resources that I use with bicep uh you can change
2:37:35
that I use with bicep uh you can change
2:37:35
that I use with bicep uh you can change things after the fact but postgres does
2:37:38
things after the fact but postgres does
2:37:38
things after the fact but postgres does have a number of constraints about what
2:37:40
have a number of constraints about what
2:37:40
have a number of constraints about what can't be changed so
2:37:42
can't be changed so
2:37:42
can't be changed so you do some experimentation and and be
2:37:45
you do some experimentation and and be
2:37:45
you do some experimentation and and be confident about your setup uh before
2:37:49
confident about your setup uh before
2:37:49
confident about your setup uh before all right other tips uh bicep the bicep
2:37:52
all right other tips uh bicep the bicep
2:37:52
all right other tips uh bicep the bicep team really wants to make uh it a nice
2:37:55
team really wants to make uh it a nice
2:37:55
team really wants to make uh it a nice experience to write bicep so they've
2:37:57
experience to write bicep so they've
2:37:57
experience to write bicep so they've developed an extension for viscode vs
2:37:59
developed an extension for viscode vs
2:38:00
developed an extension for viscode vs code which is you know a very popular
2:38:01
code which is you know a very popular
2:38:01
code which is you know a very popular IDE comes from Microsoft and this
2:38:04
IDE comes from Microsoft and this
2:38:04
IDE comes from Microsoft and this extension has syntax highlighting it has
2:38:07
extension has syntax highlighting it has
2:38:07
extension has syntax highlighting it has Snippets it has autocomplete it has
2:38:08
Snippets it has autocomplete it has
2:38:08
Snippets it has autocomplete it has linting so I really like to use it for
2:38:11
linting so I really like to use it for
2:38:11
linting so I really like to use it for the linting personally I know other
2:38:12
the linting personally I know other
2:38:12
the linting personally I know other people really like to use it for the
2:38:13
people really like to use it for the
2:38:13
people really like to use it for the Snippets I'm just not a Snippets girl uh
2:38:16
Snippets I'm just not a Snippets girl uh
2:38:16
Snippets I'm just not a Snippets girl uh but it's super helpful so definitely if
2:38:18
but it's super helpful so definitely if
2:38:18
but it's super helpful so definitely if you are a viscode user install that
2:38:20
you are a viscode user install that
2:38:20
you are a viscode user install that extension
2:38:22
extension
2:38:22
extension uh you can also do linting at the
2:38:24
uh you can also do linting at the
2:38:24
uh you can also do linting at the command line right so you can get
2:38:25
command line right so you can get
2:38:25
command line right so you can get linting an extension or you can do it in
2:38:27
linting an extension or you can do it in
2:38:27
linting an extension or you can do it in the command line so you can use AZ bicep
2:38:29
the command line so you can use AZ bicep
2:38:29
the command line so you can use AZ bicep build Dash F and you give it the path to
2:38:32
build Dash F and you give it the path to
2:38:32
build Dash F and you give it the path to the file to lint and it will go and
2:38:35
the file to lint and it will go and
2:38:35
the file to lint and it will go and it'll check it for four errors and a
2:38:38
it'll check it for four errors and a
2:38:38
it'll check it for four errors and a really good practice is to use a CI CD
2:38:40
really good practice is to use a CI CD
2:38:40
really good practice is to use a CI CD workflow in order to always check for
2:38:42
workflow in order to always check for
2:38:42
workflow in order to always check for errors so in all of my repos where I
2:38:44
errors so in all of my repos where I
2:38:44
errors so in all of my repos where I have bicep I have this uh GitHub action
2:38:47
have bicep I have this uh GitHub action
2:38:47
have bicep I have this uh GitHub action workflow that you know attempts to run
2:38:51
workflow that you know attempts to run
2:38:51
workflow that you know attempts to run the build command and if there are
2:38:53
the build command and if there are
2:38:53
the build command and if there are errors it will fail and I'll get you
2:38:55
errors it will fail and I'll get you
2:38:55
errors it will fail and I'll get you know I'll get tilt
2:38:57
know I'll get tilt
2:38:57
know I'll get tilt foreign
2:38:59
foreign
2:38:59
foreign is to not just look for errors but look
2:39:02
is to not just look for errors but look
2:39:02
is to not just look for errors but look for security issues right so you could
2:39:04
for security issues right so you could
2:39:04
for security issues right so you could have a bicep file and it's technically
2:39:06
have a bicep file and it's technically
2:39:06
have a bicep file and it's technically valid bicep but maybe you know it's
2:39:09
valid bicep but maybe you know it's
2:39:09
valid bicep but maybe you know it's using uh you know username and password
2:39:11
using uh you know username and password
2:39:11
using uh you know username and password when a better approach is to use manage
2:39:14
when a better approach is to use manage
2:39:14
when a better approach is to use manage identity so there's this new action this
2:39:17
identity so there's this new action this
2:39:17
identity so there's this new action this Microsoft security devops action that
2:39:19
Microsoft security devops action that
2:39:19
Microsoft security devops action that can analyze templates and it will let
2:39:22
can analyze templates and it will let
2:39:22
can analyze templates and it will let you know about all these potential
2:39:24
you know about all these potential
2:39:24
you know about all these potential security errors in your templates so
2:39:27
security errors in your templates so
2:39:27
security errors in your templates so we've started to add this it's very new
2:39:29
we've started to add this it's very new
2:39:29
we've started to add this it's very new so we have started to add this to our
2:39:32
so we have started to add this to our
2:39:32
so we have started to add this to our repos I have it in a few of them
2:39:34
repos I have it in a few of them
2:39:34
repos I have it in a few of them um and just working through some issues
2:39:36
um and just working through some issues
2:39:36
um and just working through some issues because we have very complicated uh
2:39:38
because we have very complicated uh
2:39:38
because we have very complicated uh bicep in in some of ours but the cool
2:39:40
bicep in in some of ours but the cool
2:39:40
bicep in in some of ours but the cool thing is like once you do it you're
2:39:42
thing is like once you do it you're
2:39:42
thing is like once you do it you're going to get this workflow and it'll you
2:39:45
going to get this workflow and it'll you
2:39:45
going to get this workflow and it'll you know it'll error it can error out if you
2:39:47
know it'll error it can error out if you
2:39:47
know it'll error it can error out if you want it to error out and it'll say like
2:39:49
want it to error out and it'll say like
2:39:49
want it to error out and it'll say like okay on this line you know you're doing
2:39:51
okay on this line you know you're doing
2:39:51
okay on this line you know you're doing this thing where you're using a password
2:39:54
this thing where you're using a password
2:39:54
this thing where you're using a password it would be better if you use manage
2:39:56
it would be better if you use manage
2:39:56
it would be better if you use manage identity instead so you can either have
2:39:58
identity instead so you can either have
2:39:58
identity instead so you can either have the workflow you know break when that
2:40:00
the workflow you know break when that
2:40:00
the workflow you know break when that happens or you can upload those errors
2:40:03
happens or you can upload those errors
2:40:03
happens or you can upload those errors uh to your code quality tab in in GitHub
2:40:06
uh to your code quality tab in in GitHub
2:40:06
uh to your code quality tab in in GitHub or in you know Azure devops or whatever
2:40:09
or in you know Azure devops or whatever
2:40:09
or in you know Azure devops or whatever and uh and then you'll be able to you
2:40:11
and uh and then you'll be able to you
2:40:11
and uh and then you'll be able to you know monitor them as security issues for
2:40:13
know monitor them as security issues for
2:40:13
know monitor them as security issues for your repo so this is another really nice
2:40:16
your repo so this is another really nice
2:40:16
your repo so this is another really nice thing to do because we don't want us to
2:40:17
thing to do because we don't want us to
2:40:17
thing to do because we don't want us to have correct bicep we want to have
2:40:19
have correct bicep we want to have
2:40:19
have correct bicep we want to have secure bicep we want to have well you
2:40:21
secure bicep we want to have well you
2:40:21
secure bicep we want to have well you want to have secure you know Azure
2:40:23
want to have secure you know Azure
2:40:23
want to have secure you know Azure resources
2:40:24
resources
2:40:24
resources uh so I just I love I you know in the
2:40:26
uh so I just I love I you know in the
2:40:27
uh so I just I love I you know in the age of you know co-pilot and chat gbt I
2:40:29
age of you know co-pilot and chat gbt I
2:40:29
age of you know co-pilot and chat gbt I love linters so any sort of linking I
2:40:31
love linters so any sort of linking I
2:40:31
love linters so any sort of linking I can do I will do it uh because I want to
2:40:33
can do I will do it uh because I want to
2:40:33
can do I will do it uh because I want to have the confidence that what I'm
2:40:35
have the confidence that what I'm
2:40:35
have the confidence that what I'm creating is you know is going to be
2:40:37
creating is you know is going to be
2:40:37
creating is you know is going to be really successful
2:40:40
uh no another thing you could do is you
2:40:42
uh no another thing you could do is you
2:40:42
uh no another thing you could do is you know I hear have I've showed postgres I
2:40:44
know I hear have I've showed postgres I
2:40:44
know I hear have I've showed postgres I use bicep just to make postgres but of
2:40:47
use bicep just to make postgres but of
2:40:47
use bicep just to make postgres but of course you can use biceps create all of
2:40:49
course you can use biceps create all of
2:40:49
course you can use biceps create all of your Azure resources and most time
2:40:50
your Azure resources and most time
2:40:50
your Azure resources and most time that's what you're going to do right
2:40:51
that's what you're going to do right
2:40:51
that's what you're going to do right you're going to be making you know your
2:40:52
you're going to be making you know your
2:40:52
you're going to be making you know your VMS and app search and all that stuff uh
2:40:54
VMS and app search and all that stuff uh
2:40:54
VMS and app search and all that stuff uh so when I do that I like to use the
2:40:56
so when I do that I like to use the
2:40:56
so when I do that I like to use the Azure developer CLI this is a new a new
2:40:59
Azure developer CLI this is a new a new
2:40:59
Azure developer CLI this is a new a new CLI tool to take care of the entire
2:41:01
CLI tool to take care of the entire
2:41:01
CLI tool to take care of the entire deployment workflow so it requires uh
2:41:04
deployment workflow so it requires uh
2:41:04
deployment workflow so it requires uh you know bicep files it also requires
2:41:06
you know bicep files it also requires
2:41:06
you know bicep files it also requires another yaml file a few other things and
2:41:09
another yaml file a few other things and
2:41:09
another yaml file a few other things and it will make it super easy to both
2:41:11
it will make it super easy to both
2:41:11
it will make it super easy to both create all your resources but also
2:41:13
create all your resources but also
2:41:13
create all your resources but also bundle up your code and send it to the
2:41:16
bundle up your code and send it to the
2:41:16
bundle up your code and send it to the server so that's that's the part I love
2:41:18
server so that's that's the part I love
2:41:18
server so that's that's the part I love of it is that it's gonna you know create
2:41:19
of it is that it's gonna you know create
2:41:19
of it is that it's gonna you know create all my resources in Azure and then you
2:41:22
all my resources in Azure and then you
2:41:22
all my resources in Azure and then you know it'll bundle up my container app
2:41:23
know it'll bundle up my container app
2:41:23
know it'll bundle up my container app and uh you know you know Docker Docker
2:41:26
and uh you know you know Docker Docker
2:41:26
and uh you know you know Docker Docker build and send that to the cloud or
2:41:27
build and send that to the cloud or
2:41:27
build and send that to the cloud or it'll one you know zip zip up my app
2:41:29
it'll one you know zip zip up my app
2:41:30
it'll one you know zip zip up my app code and and bundle and send that to the
2:41:32
code and and bundle and send that to the
2:41:32
code and and bundle and send that to the cloud so I love to use the Azure
2:41:34
cloud so I love to use the Azure
2:41:34
cloud so I love to use the Azure developer CLI to take care of everything
2:41:38
developer CLI to take care of everything
2:41:38
developer CLI to take care of everything um it also does cicd it does monitoring
2:41:41
um it also does cicd it does monitoring
2:41:41
um it also does cicd it does monitoring it does a lot so this is a new a new CLI
2:41:44
it does a lot so this is a new a new CLI
2:41:44
it does a lot so this is a new a new CLI different from the Azure CLI it's easier
2:41:46
different from the Azure CLI it's easier
2:41:46
different from the Azure CLI it's easier developer CLI but it's really it's
2:41:49
developer CLI but it's really it's
2:41:49
developer CLI but it's really it's really really cool and um I I've made 15
2:41:52
really really cool and um I I've made 15
2:41:52
really really cool and um I I've made 15 templates for it in the last you know
2:41:54
templates for it in the last you know
2:41:54
templates for it in the last you know like four months because I love it so
2:41:55
like four months because I love it so
2:41:55
like four months because I love it so much and so you can find app templates
2:41:58
much and so you can find app templates
2:41:58
much and so you can find app templates that use postgres in the azd templates
2:42:00
that use postgres in the azd templates
2:42:00
that use postgres in the azd templates Gallery uh so there's a link here and
2:42:03
Gallery uh so there's a link here and
2:42:03
Gallery uh so there's a link here and it's you know I just already filtered it
2:42:05
it's you know I just already filtered it
2:42:05
it's you know I just already filtered it by postgres for you so if you want to
2:42:07
by postgres for you so if you want to
2:42:07
by postgres for you so if you want to find examples of full applications
2:42:09
find examples of full applications
2:42:09
find examples of full applications within you know their entire bicep files
2:42:12
within you know their entire bicep files
2:42:12
within you know their entire bicep files written you can look in that azd
2:42:15
written you can look in that azd
2:42:15
written you can look in that azd templates gallery and just start from
2:42:16
templates gallery and just start from
2:42:16
templates gallery and just start from those existing ones I have written
2:42:19
those existing ones I have written
2:42:19
those existing ones I have written mostly python ones because you know
2:42:20
mostly python ones because you know
2:42:20
mostly python ones because you know that's my thing
2:42:22
that's my thing
2:42:22
that's my thing um but in terms of the infrastructure
2:42:24
um but in terms of the infrastructure
2:42:24
um but in terms of the infrastructure the infrastructure shouldn't be that
2:42:26
the infrastructure shouldn't be that
2:42:26
the infrastructure shouldn't be that difference between Python and another
2:42:27
difference between Python and another
2:42:27
difference between Python and another language like rust or go or whatever
2:42:30
language like rust or go or whatever
2:42:30
language like rust or go or whatever um so you should be able to be able to
2:42:31
um so you should be able to be able to
2:42:31
um so you should be able to be able to take the imperfiles that I've already
2:42:33
take the imperfiles that I've already
2:42:33
take the imperfiles that I've already written for a particular you know
2:42:34
written for a particular you know
2:42:34
written for a particular you know resource architecture and then use them
2:42:36
resource architecture and then use them
2:42:36
resource architecture and then use them for whatever language you're using
2:42:40
for whatever language you're using
2:42:40
for whatever language you're using foreign
2:42:42
foreign
2:42:42
foreign yeah then finally uh some more bicep
2:42:45
yeah then finally uh some more bicep
2:42:45
yeah then finally uh some more bicep writing tips because I do hear a lot
2:42:47
writing tips because I do hear a lot
2:42:47
writing tips because I do hear a lot that people don't like to write bicep
2:42:49
that people don't like to write bicep
2:42:49
that people don't like to write bicep you know it's a new language because
2:42:51
you know it's a new language because
2:42:51
you know it's a new language because it's always hard when you're you know
2:42:52
it's always hard when you're you know
2:42:52
it's always hard when you're you know writing a good language
2:42:54
writing a good language
2:42:54
writing a good language um so I always like to keep the bicep
2:42:56
um so I always like to keep the bicep
2:42:56
um so I always like to keep the bicep reference uh open uh so I'll look up
2:42:59
reference uh open uh so I'll look up
2:42:59
reference uh open uh so I'll look up whatever I'm currently working on right
2:43:00
whatever I'm currently working on right
2:43:00
whatever I'm currently working on right so the postgresql flexible servers bicep
2:43:03
so the postgresql flexible servers bicep
2:43:03
so the postgresql flexible servers bicep reference it has all the properties
2:43:04
reference it has all the properties
2:43:04
reference it has all the properties listed out so I'll just always have that
2:43:06
listed out so I'll just always have that
2:43:06
listed out so I'll just always have that open in a tab just to look at
2:43:09
open in a tab just to look at
2:43:09
open in a tab just to look at um and it also has examples at the
2:43:11
um and it also has examples at the
2:43:11
um and it also has examples at the bottom of each each page and it does
2:43:13
bottom of each each page and it does
2:43:13
bottom of each each page and it does also have examples in both arm and
2:43:15
also have examples in both arm and
2:43:15
also have examples in both arm and terraform so if you watch this but just
2:43:17
terraform so if you watch this but just
2:43:17
terraform so if you watch this but just you know determine you still want to use
2:43:18
you know determine you still want to use
2:43:18
you know determine you still want to use terraform you can use that you know the
2:43:20
terraform you can use that you know the
2:43:20
terraform you can use that you know the same reference has terraform as well
2:43:24
same reference has terraform as well
2:43:24
same reference has terraform as well uh you can search GitHub for examples
2:43:26
uh you can search GitHub for examples
2:43:26
uh you can search GitHub for examples using laying bicep I use GitHub code
2:43:29
using laying bicep I use GitHub code
2:43:29
using laying bicep I use GitHub code search all the time to help me figure
2:43:31
search all the time to help me figure
2:43:31
search all the time to help me figure out how to do something new and so I
2:43:34
out how to do something new and so I
2:43:34
out how to do something new and so I love to search for you know four
2:43:35
love to search for you know four
2:43:35
love to search for you know four examples using link biceps so I'll do
2:43:37
examples using link biceps so I'll do
2:43:37
examples using link biceps so I'll do like laying colon bicep and then
2:43:39
like laying colon bicep and then
2:43:39
like laying colon bicep and then whatever particular thing I'm looking
2:43:40
whatever particular thing I'm looking
2:43:40
whatever particular thing I'm looking for like I've searched for v-net before
2:43:42
for like I've searched for v-net before
2:43:42
for like I've searched for v-net before you know just the name of the resource
2:43:44
you know just the name of the resource
2:43:44
you know just the name of the resource and I can look at other examples see oh
2:43:46
and I can look at other examples see oh
2:43:46
and I can look at other examples see oh is somebody doing something similar to
2:43:48
is somebody doing something similar to
2:43:48
is somebody doing something similar to what I'm looking for so I think that's a
2:43:50
what I'm looking for so I think that's a
2:43:50
what I'm looking for so I think that's a great thing to do for whatever sort of
2:43:51
great thing to do for whatever sort of
2:43:51
great thing to do for whatever sort of new code you're trying to write another
2:43:54
new code you're trying to write another
2:43:54
new code you're trying to write another thing you can do is that you can export
2:43:56
thing you can do is that you can export
2:43:56
thing you can do is that you can export arm templates from the portal so if you
2:43:58
arm templates from the portal so if you
2:43:58
arm templates from the portal so if you did use you know the Porto UI to create
2:44:00
did use you know the Porto UI to create
2:44:00
did use you know the Porto UI to create a resource and exists in the portal and
2:44:02
a resource and exists in the portal and
2:44:02
a resource and exists in the portal and you're trying to figure out how to use
2:44:03
you're trying to figure out how to use
2:44:03
you're trying to figure out how to use bicep for it you can go to the portal
2:44:06
bicep for it you can go to the portal
2:44:06
bicep for it you can go to the portal and you can go on the bottom to like
2:44:09
and you can go on the bottom to like
2:44:09
and you can go on the bottom to like templates and you click like export
2:44:10
templates and you click like export
2:44:10
templates and you click like export template and you'll get the arm Json and
2:44:13
template and you'll get the arm Json and
2:44:13
template and you'll get the arm Json and then you can use AZ bicep decompile to
2:44:15
then you can use AZ bicep decompile to
2:44:15
then you can use AZ bicep decompile to turn that into bicep and uh and then you
2:44:18
turn that into bicep and uh and then you
2:44:18
turn that into bicep and uh and then you can start from there the thing to know
2:44:20
can start from there the thing to know
2:44:20
can start from there the thing to know about that is that it creates a ton like
2:44:22
about that is that it creates a ton like
2:44:22
about that is that it creates a ton like the arm jsons there will have every
2:44:24
the arm jsons there will have every
2:44:24
the arm jsons there will have every possible configuration you don't need
2:44:26
possible configuration you don't need
2:44:26
possible configuration you don't need all of it so a lot of times I'll like
2:44:28
all of it so a lot of times I'll like
2:44:28
all of it so a lot of times I'll like look at it to see you know what is in
2:44:31
look at it to see you know what is in
2:44:31
look at it to see you know what is in here and then I I cut it down because
2:44:32
here and then I I cut it down because
2:44:32
here and then I I cut it down because you don't you don't you don't need you
2:44:34
you don't you don't you don't need you
2:44:34
you don't you don't you don't need you don't need it all you just need the the
2:44:36
don't need it all you just need the the
2:44:36
don't need it all you just need the the minimal you know override configuration
2:44:38
minimal you know override configuration
2:44:38
minimal you know override configuration and I also have a blog post that talks
2:44:40
and I also have a blog post that talks
2:44:40
and I also have a blog post that talks about this more so you can read that
2:44:42
about this more so you can read that
2:44:42
about this more so you can read that blog post for for more bicep writing
2:44:44
blog post for for more bicep writing
2:44:44
blog post for for more bicep writing tips
2:44:46
tips
2:44:46
tips foreign
2:44:48
foreign
2:44:48
foreign so this is what I wanted to share with
2:44:50
so this is what I wanted to share with
2:44:50
so this is what I wanted to share with you today uh so thank you for watching I
2:44:53
you today uh so thank you for watching I
2:44:54
you today uh so thank you for watching I hope you've learned some bicep and that
2:44:55
hope you've learned some bicep and that
2:44:55
hope you've learned some bicep and that you know maybe you want to try it out
2:44:57
you know maybe you want to try it out
2:44:57
you know maybe you want to try it out yourself uh you can grab the slides at
2:45:00
yourself uh you can grab the slides at
2:45:00
yourself uh you can grab the slides at this akka dot Ms slash postgres bicep
2:45:03
this akka dot Ms slash postgres bicep
2:45:03
this akka dot Ms slash postgres bicep dot slides and uh and you can follow all
2:45:06
dot slides and uh and you can follow all
2:45:06
dot slides and uh and you can follow all the links in the slide so that you can
2:45:08
the links in the slide so that you can
2:45:08
the links in the slide so that you can take take all of the bicep code and uh
2:45:11
take take all of the bicep code and uh
2:45:11
take take all of the bicep code and uh and you know hopefully make it easier
2:45:13
and you know hopefully make it easier
2:45:13
and you know hopefully make it easier for you to get started with bicep and
2:45:15
for you to get started with bicep and
2:45:15
for you to get started with bicep and you can toot at me or tweet at me or get
2:45:17
you can toot at me or tweet at me or get
2:45:18
you can toot at me or tweet at me or get help at me or whatever Discord at me
2:45:20
help at me or whatever Discord at me
2:45:20
help at me or whatever Discord at me I'll be in the Discord after as well uh
2:45:22
I'll be in the Discord after as well uh
2:45:22
I'll be in the Discord after as well uh to let me know about your experience
2:45:24
to let me know about your experience
2:45:24
to let me know about your experience with bicep and if you're looking for any
2:45:25
with bicep and if you're looking for any
2:45:25
with bicep and if you're looking for any particular examples I'm always looking
2:45:27
particular examples I'm always looking
2:45:27
particular examples I'm always looking for excuses to you know try something
2:45:29
for excuses to you know try something
2:45:29
for excuses to you know try something out So yeah thank you for coming today
2:45:32
out So yeah thank you for coming today
2:45:32
out So yeah thank you for coming today that is my talk
2:45:35
that is my talk
2:45:35
that is my talk excellent uh I have to say like yeah
2:45:37
excellent uh I have to say like yeah
2:45:37
excellent uh I have to say like yeah that was really great I've learned quite
2:45:39
that was really great I've learned quite
2:45:39
that was really great I've learned quite a bit about bicep I did not really know
2:45:41
a bit about bicep I did not really know
2:45:41
a bit about bicep I did not really know that much about bicep it seems like it
2:45:43
that much about bicep it seems like it
2:45:43
that much about bicep it seems like it is one of these like uh I guess next
2:45:45
is one of these like uh I guess next
2:45:45
is one of these like uh I guess next Generation like you said like terraform
2:45:47
Generation like you said like terraform
2:45:47
Generation like you said like terraform or like palumi uh so it kind of Falls in
2:45:50
or like palumi uh so it kind of Falls in
2:45:50
or like palumi uh so it kind of Falls in that category so it's really good
2:45:52
that category so it's really good
2:45:52
that category so it's really good um I I know we're a bit up on the clock
2:45:55
um I I know we're a bit up on the clock
2:45:55
um I I know we're a bit up on the clock but I did want to ask you a quick
2:45:56
but I did want to ask you a quick
2:45:56
but I did want to ask you a quick question like if I'm deploying postgres
2:45:57
question like if I'm deploying postgres
2:45:57
question like if I'm deploying postgres I'm probably going to be needing to
2:45:59
I'm probably going to be needing to
2:45:59
I'm probably going to be needing to deploy PG bouncer with it uh is that
2:46:02
deploy PG bouncer with it uh is that
2:46:02
deploy PG bouncer with it uh is that easy or is that difficult if I want to
2:46:04
easy or is that difficult if I want to
2:46:04
easy or is that difficult if I want to do that with bicep
2:46:05
do that with bicep
2:46:05
do that with bicep uh well it is your PT bouncer is enabled
2:46:08
uh well it is your PT bouncer is enabled
2:46:08
uh well it is your PT bouncer is enabled by default on the postgres server so you
2:46:11
by default on the postgres server so you
2:46:11
by default on the postgres server so you just got it on Azure database for
2:46:13
just got it on Azure database for
2:46:13
just got it on Azure database for postgres you mean on Flex server yeah
2:46:15
postgres you mean on Flex server yeah
2:46:15
postgres you mean on Flex server yeah sorry to be very specific the one that I
2:46:16
sorry to be very specific the one that I
2:46:16
sorry to be very specific the one that I showed is enabled by default uh is it do
2:46:19
showed is enabled by default uh is it do
2:46:19
showed is enabled by default uh is it do you know about with cytus
2:46:23
you know about with cytus
2:46:23
you know about with cytus I imagine it must work pretty similarly
2:46:25
I imagine it must work pretty similarly
2:46:25
I imagine it must work pretty similarly then yeah so it sounds like it'd be
2:46:26
then yeah so it sounds like it'd be
2:46:26
then yeah so it sounds like it'd be pretty easy
2:46:27
pretty easy
2:46:27
pretty easy yeah yeah so actually the ones I showed
2:46:29
yeah yeah so actually the ones I showed
2:46:29
yeah yeah so actually the ones I showed it's just already enabled and you can go
2:46:31
it's just already enabled and you can go
2:46:31
it's just already enabled and you can go into server parameters in the portal to
2:46:32
into server parameters in the portal to
2:46:33
into server parameters in the portal to to look at all those settings but uh
2:46:34
to look at all those settings but uh
2:46:34
to look at all those settings but uh it's it's already enabled okay well and
2:46:36
it's it's already enabled okay well and
2:46:36
it's it's already enabled okay well and I know I know Yalta is on the Discord um
2:46:39
I know I know Yalta is on the Discord um
2:46:39
I know I know Yalta is on the Discord um chat and he's both on the status project
2:46:42
chat and he's both on the status project
2:46:42
chat and he's both on the status project and a PG bouncer maintainer so if
2:46:44
and a PG bouncer maintainer so if
2:46:44
and a PG bouncer maintainer so if anybody has any questions about that
2:46:46
anybody has any questions about that
2:46:46
anybody has any questions about that they can definitely pop them in on the
2:46:47
they can definitely pop them in on the
2:46:47
they can definitely pop them in on the Discord Pamela thank you so much for
2:46:50
Discord Pamela thank you so much for
2:46:50
Discord Pamela thank you so much for being being part of cytuscon today I
2:46:53
being being part of cytuscon today I
2:46:53
being being part of cytuscon today I also learned something about the naming
2:46:54
also learned something about the naming
2:46:54
also learned something about the naming of bicep and that there's a connection
2:46:55
of bicep and that there's a connection
2:46:55
of bicep and that there's a connection between bicep and arm okay I didn't
2:46:59
between bicep and arm okay I didn't
2:46:59
between bicep and arm okay I didn't actually connect those dots beforehand
2:47:01
actually connect those dots beforehand
2:47:01
actually connect those dots beforehand so
2:47:02
so
2:47:02
so yeah yeah I literally only connected
2:47:04
yeah yeah I literally only connected
2:47:04
yeah yeah I literally only connected them this morning
2:47:07
them this morning
2:47:07
them this morning all right well thank you so much for
2:47:09
all right well thank you so much for
2:47:09
all right well thank you so much for being here appreciate it and you're
2:47:11
being here appreciate it and you're
2:47:11
being here appreciate it and you're gonna go to the Discord to answer
2:47:12
gonna go to the Discord to answer
2:47:12
gonna go to the Discord to answer questions and chit chat all right
2:47:14
questions and chit chat all right
2:47:14
questions and chit chat all right awesome excellent all right uh and with
2:47:17
awesome excellent all right uh and with
2:47:17
awesome excellent all right uh and with that let us then bring in Melanie
2:47:20
that let us then bring in Melanie
2:47:20
that let us then bring in Melanie plagueman uh she is helping us uh sort
2:47:23
plagueman uh she is helping us uh sort
2:47:23
plagueman uh she is helping us uh sort of round out our third period of uh the
2:47:26
of round out our third period of uh the
2:47:26
of round out our third period of uh the show today uh so excellent excellent uh
2:47:29
show today uh so excellent excellent uh
2:47:29
show today uh so excellent excellent uh how you doing good how are you
2:47:31
how you doing good how are you
2:47:31
how you doing good how are you I'm doing well it's been good so far for
2:47:34
I'm doing well it's been good so far for
2:47:34
I'm doing well it's been good so far for people who don't know Melanie um Melanie
2:47:37
people who don't know Melanie um Melanie
2:47:37
people who don't know Melanie um Melanie is a senior software engineer
2:47:39
is a senior software engineer
2:47:39
is a senior software engineer um at Microsoft and she's a postgres
2:47:41
um at Microsoft and she's a postgres
2:47:41
um at Microsoft and she's a postgres hacker and postgres contributor
2:47:44
hacker and postgres contributor
2:47:44
hacker and postgres contributor and just uh for those who don't pay
2:47:46
and just uh for those who don't pay
2:47:46
and just uh for those who don't pay attention to the postgres commits which
2:47:47
attention to the postgres commits which
2:47:48
attention to the postgres commits which maybe is a larger number of people uh
2:47:50
maybe is a larger number of people uh
2:47:50
maybe is a larger number of people uh than you might expect she recently
2:47:52
than you might expect she recently
2:47:52
than you might expect she recently contributed PG statio into postgres 16
2:47:55
contributed PG statio into postgres 16
2:47:55
contributed PG statio into postgres 16 uh fingers crossed that will be
2:47:56
uh fingers crossed that will be
2:47:56
uh fingers crossed that will be available later this year uh and I guess
2:47:59
available later this year uh and I guess
2:47:59
available later this year uh and I guess that's a big part of what you're going
2:48:00
that's a big part of what you're going
2:48:00
that's a big part of what you're going to talk about today yep that is
2:48:02
to talk about today yep that is
2:48:02
to talk about today yep that is observability and postgres and I know
2:48:04
observability and postgres and I know
2:48:04
observability and postgres and I know people are really excited about PG
2:48:06
people are really excited about PG
2:48:06
people are really excited about PG statio so let's find out why take it
2:48:08
statio so let's find out why take it
2:48:08
statio so let's find out why take it away Melanie
2:48:13
yeah so this talk is going to be great
2:48:15
yeah so this talk is going to be great
2:48:15
yeah so this talk is going to be great for users that are already comfortable
2:48:18
for users that are already comfortable
2:48:18
for users that are already comfortable doing some amount of tuning and want to
2:48:20
doing some amount of tuning and want to
2:48:20
doing some amount of tuning and want to kind of take it to the next level and so
2:48:23
kind of take it to the next level and so
2:48:23
kind of take it to the next level and so that's really who it's targeted at
2:48:27
so uh I'm going to talk about the
2:48:30
so uh I'm going to talk about the
2:48:30
so uh I'm going to talk about the difference between sort of the existing
2:48:32
difference between sort of the existing
2:48:32
difference between sort of the existing i o statistics that are available in
2:48:34
i o statistics that are available in
2:48:34
i o statistics that are available in postgres and what PG statio can offer in
2:48:37
postgres and what PG statio can offer in
2:48:37
postgres and what PG statio can offer in terms of observability
2:48:40
terms of observability
2:48:40
terms of observability just about me I'm a postgres hacker
2:48:43
just about me I'm a postgres hacker
2:48:43
just about me I'm a postgres hacker working at Microsoft on the open source
2:48:45
working at Microsoft on the open source
2:48:45
working at Microsoft on the open source postgres team the last two years I've
2:48:48
postgres team the last two years I've
2:48:48
postgres team the last two years I've been very focused on i o performance and
2:48:50
been very focused on i o performance and
2:48:50
been very focused on i o performance and I O benchmarking but before that I've
2:48:52
I O benchmarking but before that I've
2:48:52
I O benchmarking but before that I've worked on
2:48:54
worked on
2:48:54
worked on um a lot of the other subsystems and
2:48:56
um a lot of the other subsystems and
2:48:56
um a lot of the other subsystems and postgres like planner and executor and
2:48:58
postgres like planner and executor and
2:48:58
postgres like planner and executor and things like that
2:49:00
things like that
2:49:00
things like that um
2:49:01
um
2:49:01
um so users have pretty much two main goals
2:49:05
so users have pretty much two main goals
2:49:05
so users have pretty much two main goals for their transactional workload i o
2:49:07
for their transactional workload i o
2:49:07
for their transactional workload i o performance they care about throughput
2:49:09
performance they care about throughput
2:49:09
performance they care about throughput and they care about latency right so you
2:49:11
and they care about latency right so you
2:49:11
and they care about latency right so you want High TPS and low latency low tail
2:49:15
want High TPS and low latency low tail
2:49:15
want High TPS and low latency low tail latency
2:49:16
latency
2:49:16
latency um that's when you're thinking about
2:49:18
um that's when you're thinking about
2:49:18
um that's when you're thinking about performance those are the things that
2:49:20
performance those are the things that
2:49:20
performance those are the things that that really matter you can boil
2:49:21
that really matter you can boil
2:49:21
that really matter you can boil everything down to that but some of the
2:49:24
everything down to that but some of the
2:49:24
everything down to that but some of the impediments to that I mean the most sort
2:49:27
impediments to that I mean the most sort
2:49:27
impediments to that I mean the most sort of obvious one is if your data is not in
2:49:29
of obvious one is if your data is not in
2:49:29
of obvious one is if your data is not in shared buffers so if your data is either
2:49:33
shared buffers so if your data is either
2:49:33
shared buffers so if your data is either too large for shared buffers and you
2:49:36
too large for shared buffers and you
2:49:36
too large for shared buffers and you don't expect it to fit or you've
2:49:38
don't expect it to fit or you've
2:49:38
don't expect it to fit or you've misconfigured shared buffers then you're
2:49:40
misconfigured shared buffers then you're
2:49:40
misconfigured shared buffers then you're going to see more IO and that can be a
2:49:42
going to see more IO and that can be a
2:49:42
going to see more IO and that can be a common performance issue
2:49:45
common performance issue
2:49:45
common performance issue um the cause of common performance issue
2:49:47
um the cause of common performance issue
2:49:47
um the cause of common performance issue uh another one is issues around wall i o
2:49:50
uh another one is issues around wall i o
2:49:50
uh another one is issues around wall i o and checkpoint or misconfiguration
2:49:53
and checkpoint or misconfiguration
2:49:53
and checkpoint or misconfiguration um that's a pretty common reason that
2:49:54
um that's a pretty common reason that
2:49:54
um that's a pretty common reason that people are seeing IO bottlenecks and
2:49:57
people are seeing IO bottlenecks and
2:49:57
people are seeing IO bottlenecks and then Auto vacuum tuning auto vacuum is
2:49:59
then Auto vacuum tuning auto vacuum is
2:49:59
then Auto vacuum tuning auto vacuum is notoriously finicky and it really
2:50:01
notoriously finicky and it really
2:50:01
notoriously finicky and it really depends on your workload and the out of
2:50:03
depends on your workload and the out of
2:50:03
depends on your workload and the out of the box settings are going to be good
2:50:04
the box settings are going to be good
2:50:04
the box settings are going to be good for No One basically so
2:50:08
for No One basically so
2:50:08
for No One basically so um you know if your auto vacuum is not
2:50:09
um you know if your auto vacuum is not
2:50:09
um you know if your auto vacuum is not running frequently enough if you're not
2:50:13
running frequently enough if you're not
2:50:13
running frequently enough if you're not um if there's uh the workers aren't
2:50:15
um if there's uh the workers aren't
2:50:15
um if there's uh the workers aren't being aggressive enough that kind of
2:50:16
being aggressive enough that kind of
2:50:16
being aggressive enough that kind of thing
2:50:17
thing
2:50:17
thing so
2:50:18
so
2:50:18
so as we think about tuning the these kind
2:50:21
as we think about tuning the these kind
2:50:21
as we think about tuning the these kind of are bucketed into different areas so
2:50:24
of are bucketed into different areas so
2:50:24
of are bucketed into different areas so you can tune shared buffers there are
2:50:26
you can tune shared buffers there are
2:50:26
you can tune shared buffers there are different postgres skux around the
2:50:29
different postgres skux around the
2:50:29
different postgres skux around the background writer around checkpointer
2:50:31
background writer around checkpointer
2:50:31
background writer around checkpointer and wall tuning and around Auto vacuum
2:50:34
and wall tuning and around Auto vacuum
2:50:34
and wall tuning and around Auto vacuum I'm not gonna get too much into
2:50:36
I'm not gonna get too much into
2:50:36
I'm not gonna get too much into checkpoint or an auto vacuum I'm
2:50:37
checkpoint or an auto vacuum I'm
2:50:37
checkpoint or an auto vacuum I'm primarily going to focus on what sort of
2:50:40
primarily going to focus on what sort of
2:50:40
primarily going to focus on what sort of data we can use to indicate if we need
2:50:44
data we can use to indicate if we need
2:50:44
data we can use to indicate if we need to make a change to Shared buffers or
2:50:46
to make a change to Shared buffers or
2:50:46
to make a change to Shared buffers or it's a background writer configuration
2:50:50
and as we think about tuning you know
2:50:54
and as we think about tuning you know
2:50:54
and as we think about tuning you know the ideal is that you're data driven so
2:50:56
the ideal is that you're data driven so
2:50:56
the ideal is that you're data driven so you want a base tuning decisions on
2:50:59
you want a base tuning decisions on
2:50:59
you want a base tuning decisions on actually what's happening in the
2:51:02
actually what's happening in the
2:51:02
actually what's happening in the database and so how do you get that
2:51:03
database and so how do you get that
2:51:03
database and so how do you get that information
2:51:05
information
2:51:05
information um so one option is to use operating
2:51:08
um so one option is to use operating
2:51:08
um so one option is to use operating system statistics and utilities of like
2:51:11
system statistics and utilities of like
2:51:11
system statistics and utilities of like using iOS that and other
2:51:13
using iOS that and other
2:51:13
using iOS that and other um there's a you know different
2:51:14
um there's a you know different
2:51:14
um there's a you know different extensions in postgres that'll give you
2:51:16
extensions in postgres that'll give you
2:51:16
extensions in postgres that'll give you access to some operating system
2:51:18
access to some operating system
2:51:18
access to some operating system information about IO
2:51:20
information about IO
2:51:20
information about IO um but the kind of the first place the
2:51:22
um but the kind of the first place the
2:51:22
um but the kind of the first place the most obvious place that you would go is
2:51:24
most obvious place that you would go is
2:51:24
most obvious place that you would go is the existing i o statistics in postgres
2:51:26
the existing i o statistics in postgres
2:51:26
the existing i o statistics in postgres so uh PG stats database is going to give
2:51:30
so uh PG stats database is going to give
2:51:30
so uh PG stats database is going to give you per database statistics
2:51:32
you per database statistics
2:51:32
you per database statistics um on reads and writes hits and
2:51:35
um on reads and writes hits and
2:51:35
um on reads and writes hits and um sorry read and write time hits and
2:51:37
um sorry read and write time hits and
2:51:37
um sorry read and write time hits and reads uh there's PG statio tables and
2:51:40
reads uh there's PG statio tables and
2:51:40
reads uh there's PG statio tables and indexes there's pgs.bg writer and then
2:51:43
indexes there's pgs.bg writer and then
2:51:43
indexes there's pgs.bg writer and then many many users are using the PG stat
2:51:45
many many users are using the PG stat
2:51:45
many many users are using the PG stat statements extension which is going to
2:51:47
statements extension which is going to
2:51:47
statements extension which is going to give you a fair amount of information
2:51:49
give you a fair amount of information
2:51:49
give you a fair amount of information but we wanted to add PG statio because
2:51:52
but we wanted to add PG statio because
2:51:52
but we wanted to add PG statio because there's actually substantial gaps in
2:51:55
there's actually substantial gaps in
2:51:55
there's actually substantial gaps in these existing i o statistics that are
2:51:57
these existing i o statistics that are
2:51:57
these existing i o statistics that are in postgres and even those that are in
2:51:59
in postgres and even those that are in
2:51:59
in postgres and even those that are in extensions and so what we really needed
2:52:02
extensions and so what we really needed
2:52:02
extensions and so what we really needed to do was actually make changes
2:52:05
to do was actually make changes
2:52:05
to do was actually make changes uh to where we're collecting statistics
2:52:08
uh to where we're collecting statistics
2:52:08
uh to where we're collecting statistics and not just what we're
2:52:10
and not just what we're
2:52:10
and not just what we're um revealing and bubbling up but
2:52:11
um revealing and bubbling up but
2:52:11
um revealing and bubbling up but actually what are we actually collecting
2:52:14
actually what are we actually collecting
2:52:14
actually what are we actually collecting and so we focused I I think you can kind
2:52:17
and so we focused I I think you can kind
2:52:17
and so we focused I I think you can kind of think of the three areas or the three
2:52:19
of think of the three areas or the three
2:52:19
of think of the three areas or the three gaps that um PG statio is addressing uh
2:52:23
gaps that um PG statio is addressing uh
2:52:23
gaps that um PG statio is addressing uh and we can kind of break those down so I
2:52:25
and we can kind of break those down so I
2:52:25
and we can kind of break those down so I think one of the most important ones is
2:52:27
think one of the most important ones is
2:52:27
think one of the most important ones is that in all of the existing i o
2:52:30
that in all of the existing i o
2:52:30
that in all of the existing i o statistics rights are include flushes
2:52:34
statistics rights are include flushes
2:52:34
statistics rights are include flushes and extends and I'm going to talk a
2:52:36
and extends and I'm going to talk a
2:52:36
and extends and I'm going to talk a little bit more about okay what is that
2:52:38
little bit more about okay what is that
2:52:38
little bit more about okay what is that anyway and why does it matter
2:52:41
anyway and why does it matter
2:52:41
anyway and why does it matter um another uh sort of the granularity is
2:52:45
um another uh sort of the granularity is
2:52:45
um another uh sort of the granularity is one of the main problems with the
2:52:46
one of the main problems with the
2:52:46
one of the main problems with the existing i o statistics so all back end
2:52:49
existing i o statistics so all back end
2:52:49
existing i o statistics so all back end types whether that's checkpoint or
2:52:51
types whether that's checkpoint or
2:52:51
types whether that's checkpoint or background writer you know whatever type
2:52:55
background writer you know whatever type
2:52:55
background writer you know whatever type of back end it is those are all going to
2:52:56
of back end it is those are all going to
2:52:57
of back end it is those are all going to be included in the same statistics
2:52:58
be included in the same statistics
2:52:58
be included in the same statistics you're not going to see it broken down
2:53:00
you're not going to see it broken down
2:53:00
you're not going to see it broken down there's some amount of uh there's some
2:53:03
there's some amount of uh there's some
2:53:03
there's some amount of uh there's some exceptions to that but largely you're
2:53:05
exceptions to that but largely you're
2:53:05
exceptions to that but largely you're going to see it all together and then
2:53:07
going to see it all together and then
2:53:07
going to see it all together and then also you're going to CIO for
2:53:09
also you're going to CIO for
2:53:09
also you're going to CIO for all different contexts in context here
2:53:12
all different contexts in context here
2:53:12
all different contexts in context here what I mean is okay vacuuming is
2:53:15
what I mean is okay vacuuming is
2:53:15
what I mean is okay vacuuming is different than uh just normal
2:53:18
different than uh just normal
2:53:18
different than uh just normal transactional workload IO uh doing a
2:53:22
transactional workload IO uh doing a
2:53:22
transactional workload IO uh doing a bulk read load doing bulk data loading
2:53:24
bulk read load doing bulk data loading
2:53:24
bulk read load doing bulk data loading that's different sort of a different i o
2:53:26
that's different sort of a different i o
2:53:26
that's different sort of a different i o pattern a different reason for doing I O
2:53:29
pattern a different reason for doing I O
2:53:29
pattern a different reason for doing I O and so you're going to want to address
2:53:31
and so you're going to want to address
2:53:31
and so you're going to want to address that differently when it comes to tuning
2:53:35
that differently when it comes to tuning
2:53:35
that differently when it comes to tuning so this is just a snapshot probably
2:53:38
so this is just a snapshot probably
2:53:38
so this is just a snapshot probably right after a little bit after I did an
2:53:40
right after a little bit after I did an
2:53:40
right after a little bit after I did an its UB so you can see a lot of it's zero
2:53:42
its UB so you can see a lot of it's zero
2:53:42
its UB so you can see a lot of it's zero but this is what Fiji setio looks like
2:53:45
but this is what Fiji setio looks like
2:53:45
but this is what Fiji setio looks like um and you can see that we actually have
2:53:48
um and you can see that we actually have
2:53:48
um and you can see that we actually have read and write and fsync uh time so
2:53:52
read and write and fsync uh time so
2:53:52
read and write and fsync uh time so that's cool
2:53:53
that's cool
2:53:53
that's cool um we can see the timing and then uh we
2:53:57
um we can see the timing and then uh we
2:53:57
um we can see the timing and then uh we have some specific columns that are
2:53:59
have some specific columns that are
2:53:59
have some specific columns that are actually relevant for buffer access
2:54:01
actually relevant for buffer access
2:54:01
actually relevant for buffer access strategies which I'll talk about a
2:54:02
strategies which I'll talk about a
2:54:02
strategies which I'll talk about a little bit later like reuses
2:54:05
little bit later like reuses
2:54:05
little bit later like reuses um so that's just to give you a little
2:54:08
um so that's just to give you a little
2:54:08
um so that's just to give you a little picture of it
2:54:10
picture of it
2:54:10
picture of it um so let's go back to what I said about
2:54:12
um so let's go back to what I said about
2:54:12
um so let's go back to what I said about what are the gaps right so why do we
2:54:15
what are the gaps right so why do we
2:54:15
what are the gaps right so why do we care about counting flushes and extends
2:54:17
care about counting flushes and extends
2:54:17
care about counting flushes and extends separately what is a flush so in PG
2:54:20
separately what is a flush so in PG
2:54:20
separately what is a flush so in PG statio we call flushes rights because
2:54:22
statio we call flushes rights because
2:54:22
statio we call flushes rights because we're able to distinguish rights from
2:54:24
we're able to distinguish rights from
2:54:24
we're able to distinguish rights from extends
2:54:25
extends
2:54:25
extends um and the what in order to kind of talk
2:54:29
um and the what in order to kind of talk
2:54:29
um and the what in order to kind of talk about this what I want to do is walk you
2:54:32
about this what I want to do is walk you
2:54:32
about this what I want to do is walk you through from an internals perspective
2:54:34
through from an internals perspective
2:54:34
through from an internals perspective the update or insert the some of the
2:54:37
the update or insert the some of the
2:54:37
the update or insert the some of the steps are pretty similar workflow in
2:54:40
steps are pretty similar workflow in
2:54:40
steps are pretty similar workflow in postgres to give you an idea of like
2:54:42
postgres to give you an idea of like
2:54:42
postgres to give you an idea of like what is different about an extend and
2:54:44
what is different about an extend and
2:54:44
what is different about an extend and why does it matter if it's separate and
2:54:46
why does it matter if it's separate and
2:54:46
why does it matter if it's separate and what does that tell us so when you are
2:54:49
what does that tell us so when you are
2:54:49
what does that tell us so when you are let's say you're going to just do an
2:54:50
let's say you're going to just do an
2:54:50
let's say you're going to just do an insert insert some data
2:54:52
insert insert some data
2:54:52
insert insert some data first you have to find a place to put it
2:54:54
first you have to find a place to put it
2:54:54
first you have to find a place to put it so your file for your relation is going
2:54:57
so your file for your relation is going
2:54:57
so your file for your relation is going to have
2:54:58
to have
2:54:58
to have you know blocks in it and so you need to
2:55:01
you know blocks in it and so you need to
2:55:01
you know blocks in it and so you need to find a block that has some space
2:55:03
find a block that has some space
2:55:03
find a block that has some space available
2:55:04
available
2:55:04
available and so that could be in the middle it
2:55:07
and so that could be in the middle it
2:55:07
and so that could be in the middle it could be at the end you're just looking
2:55:09
could be at the end you're just looking
2:55:09
could be at the end you're just looking for a block that has space uh uh for
2:55:11
for a block that has space uh uh for
2:55:11
for a block that has space uh uh for that data you're going to add and if
2:55:15
that data you're going to add and if
2:55:15
that data you're going to add and if there isn't a block that already has
2:55:16
there isn't a block that already has
2:55:16
there isn't a block that already has space you're going to extend need to
2:55:18
space you're going to extend need to
2:55:18
space you're going to extend need to extend the file to add another block to
2:55:20
extend the file to add another block to
2:55:20
extend the file to add another block to the file and if you think about it I
2:55:23
the file and if you think about it I
2:55:23
the file and if you think about it I mean you're gonna this is inevitable you
2:55:25
mean you're gonna this is inevitable you
2:55:25
mean you're gonna this is inevitable you can't avoid this if you have more data
2:55:27
can't avoid this if you have more data
2:55:27
can't avoid this if you have more data eventually you're gonna have to make the
2:55:28
eventually you're gonna have to make the
2:55:28
eventually you're gonna have to make the file bigger right so
2:55:31
file bigger right so
2:55:31
file bigger right so now that you've sort of identified which
2:55:34
now that you've sort of identified which
2:55:34
now that you've sort of identified which block you want to actually in the file
2:55:36
block you want to actually in the file
2:55:36
block you want to actually in the file you want to actually add your data to
2:55:38
you want to actually add your data to
2:55:38
you want to actually add your data to you have to get it into shared buffers
2:55:41
you have to get it into shared buffers
2:55:41
you have to get it into shared buffers so first you'll check okay is it already
2:55:43
so first you'll check okay is it already
2:55:43
so first you'll check okay is it already insured buffers if it's already insured
2:55:46
insured buffers if it's already insured
2:55:46
insured buffers if it's already insured buffers that's a cash hit right I'm sure
2:55:48
buffers that's a cash hit right I'm sure
2:55:48
buffers that's a cash hit right I'm sure buffers hit and we're done so that
2:55:50
buffers hit and we're done so that
2:55:50
buffers hit and we're done so that counts as a hit we don't do we need to
2:55:52
counts as a hit we don't do we need to
2:55:52
counts as a hit we don't do we need to do the read and we can just add our data
2:55:55
do the read and we can just add our data
2:55:55
do the read and we can just add our data however if the block is not already in
2:55:59
however if the block is not already in
2:55:59
however if the block is not already in shared buffers then we need to find a
2:56:01
shared buffers then we need to find a
2:56:01
shared buffers then we need to find a shared buffer to put it in
2:56:03
shared buffer to put it in
2:56:03
shared buffer to put it in and the shared buffer that
2:56:07
and the shared buffer that
2:56:07
and the shared buffer that in order to find a shared buffer to put
2:56:09
in order to find a shared buffer to put
2:56:09
in order to find a shared buffer to put it in we might find one and then
2:56:10
it in we might find one and then
2:56:10
it in we might find one and then actually that shirt buffer is dirty it
2:56:12
actually that shirt buffer is dirty it
2:56:12
actually that shirt buffer is dirty it has unrelated data in it that is not you
2:56:16
has unrelated data in it that is not you
2:56:16
has unrelated data in it that is not you know not necessarily data from our table
2:56:18
know not necessarily data from our table
2:56:18
know not necessarily data from our table but we can't just throw it away because
2:56:20
but we can't just throw it away because
2:56:20
but we can't just throw it away because you know then we would lose the data so
2:56:23
you know then we would lose the data so
2:56:23
you know then we would lose the data so before we can use that buffer we have to
2:56:26
before we can use that buffer we have to
2:56:26
before we can use that buffer we have to write that data out and that's a flush
2:56:28
write that data out and that's a flush
2:56:29
write that data out and that's a flush and that will count as a right MPG
2:56:31
and that will count as a right MPG
2:56:31
and that will count as a right MPG statio and you can see that here if that
2:56:34
statio and you can see that here if that
2:56:34
statio and you can see that here if that buffer had already been free or clean
2:56:36
buffer had already been free or clean
2:56:36
buffer had already been free or clean that it hadn't had dirty data in it then
2:56:39
that it hadn't had dirty data in it then
2:56:39
that it hadn't had dirty data in it then we wouldn't have needed to do this right
2:56:40
we wouldn't have needed to do this right
2:56:40
we wouldn't have needed to do this right so it's an avoidable right
2:56:43
so it's an avoidable right
2:56:43
so it's an avoidable right finally we're going to read our block
2:56:45
finally we're going to read our block
2:56:45
finally we're going to read our block into the shared buffer and now we have
2:56:48
into the shared buffer and now we have
2:56:48
into the shared buffer and now we have it and we can do our insert and this
2:56:50
it and we can do our insert and this
2:56:51
it and we can do our insert and this right is not a right like oh it's a
2:56:52
right is not a right like oh it's a
2:56:52
right is not a right like oh it's a write out to to even to Kernel buffers
2:56:55
write out to to even to Kernel buffers
2:56:55
write out to to even to Kernel buffers or disk this is just basically copying
2:56:57
or disk this is just basically copying
2:56:57
or disk this is just basically copying in our dirty data art Tuple into that
2:57:00
in our dirty data art Tuple into that
2:57:00
in our dirty data art Tuple into that buffer
2:57:02
buffer
2:57:02
buffer so what does this tell us about why it's
2:57:04
so what does this tell us about why it's
2:57:04
so what does this tell us about why it's important to count flushes and extend
2:57:06
important to count flushes and extend
2:57:06
important to count flushes and extend separately
2:57:08
separately
2:57:08
separately what we saw was that the flush of the
2:57:11
what we saw was that the flush of the
2:57:11
what we saw was that the flush of the dirty data
2:57:12
dirty data
2:57:12
dirty data from an unrelated Source was actually
2:57:15
from an unrelated Source was actually
2:57:15
from an unrelated Source was actually avoidable if we had had a clean buffer
2:57:17
avoidable if we had had a clean buffer
2:57:17
avoidable if we had had a clean buffer we wouldn't have needed to do it whereas
2:57:19
we wouldn't have needed to do it whereas
2:57:19
we wouldn't have needed to do it whereas the extend is something that is
2:57:21
the extend is something that is
2:57:21
the extend is something that is avoidable I mean is unavoidable we're
2:57:23
avoidable I mean is unavoidable we're
2:57:23
avoidable I mean is unavoidable we're going to have to do extends as the file
2:57:25
going to have to do extends as the file
2:57:25
going to have to do extends as the file gets bigger it's a consequence of our
2:57:27
gets bigger it's a consequence of our
2:57:27
gets bigger it's a consequence of our workload and not cleaning up someone
2:57:29
workload and not cleaning up someone
2:57:29
workload and not cleaning up someone else's mess basically so by separating
2:57:32
else's mess basically so by separating
2:57:32
else's mess basically so by separating them we can allow ourselves to
2:57:34
them we can allow ourselves to
2:57:34
them we can allow ourselves to understand whether or not we actually
2:57:36
understand whether or not we actually
2:57:36
understand whether or not we actually have to tune something it we could see a
2:57:40
have to tune something it we could see a
2:57:40
have to tune something it we could see a lot of Rights and say oh we probably
2:57:41
lot of Rights and say oh we probably
2:57:42
lot of Rights and say oh we probably need to increase share buffers but
2:57:43
need to increase share buffers but
2:57:43
need to increase share buffers but actually we're just doing a lot of
2:57:44
actually we're just doing a lot of
2:57:44
actually we're just doing a lot of extends and this is something you might
2:57:47
extends and this is something you might
2:57:47
extends and this is something you might see for example if you're doing a bulk
2:57:49
see for example if you're doing a bulk
2:57:49
see for example if you're doing a bulk right you're doing a copy from you might
2:57:52
right you're doing a copy from you might
2:57:52
right you're doing a copy from you might see a ton of in existing i o statistics
2:57:55
see a ton of in existing i o statistics
2:57:55
see a ton of in existing i o statistics a ton of Rights but actually these are
2:57:57
a ton of Rights but actually these are
2:57:57
a ton of Rights but actually these are just copy from doing a bunch of extents
2:57:59
just copy from doing a bunch of extents
2:57:59
just copy from doing a bunch of extents and this is not even data that's really
2:58:01
and this is not even data that's really
2:58:01
and this is not even data that's really part of your working set for your
2:58:03
part of your working set for your
2:58:03
part of your working set for your transactional workload necessarily so
2:58:06
transactional workload necessarily so
2:58:06
transactional workload necessarily so tuning for that is going to be to
2:58:09
tuning for that is going to be to
2:58:09
tuning for that is going to be to disadvantageous
2:58:12
disadvantageous
2:58:12
disadvantageous I also mentioned earlier that we want to
2:58:15
I also mentioned earlier that we want to
2:58:15
I also mentioned earlier that we want to separate out IO by the context in which
2:58:19
separate out IO by the context in which
2:58:19
separate out IO by the context in which it's done and by back-end type
2:58:22
it's done and by back-end type
2:58:22
it's done and by back-end type and to talk about that and why that's
2:58:24
and to talk about that and why that's
2:58:24
and to talk about that and why that's important I'm going to use as an example
2:58:26
important I'm going to use as an example
2:58:26
important I'm going to use as an example Auto vacuum i o workflow roughly so Auto
2:58:31
Auto vacuum i o workflow roughly so Auto
2:58:31
Auto vacuum i o workflow roughly so Auto vacuum and vacuum but let's just
2:58:33
vacuum and vacuum but let's just
2:58:33
vacuum and vacuum but let's just specifically talk about Auto vacuum is
2:58:36
specifically talk about Auto vacuum is
2:58:36
specifically talk about Auto vacuum is going to identify okay what are the
2:58:38
going to identify okay what are the
2:58:38
going to identify okay what are the blocks that I need to vacuum and then go
2:58:40
blocks that I need to vacuum and then go
2:58:40
blocks that I need to vacuum and then go through and do each one so it might not
2:58:42
through and do each one so it might not
2:58:42
through and do each one so it might not be all the blocks in the relation of
2:58:44
be all the blocks in the relation of
2:58:44
be all the blocks in the relation of course so uh it identifies the next
2:58:47
course so uh it identifies the next
2:58:47
course so uh it identifies the next block to vacuum and then it has to look
2:58:49
block to vacuum and then it has to look
2:58:49
block to vacuum and then it has to look for it if see if it's in shared buffers
2:58:51
for it if see if it's in shared buffers
2:58:51
for it if see if it's in shared buffers if it's insured buffers then it can
2:58:53
if it's insured buffers then it can
2:58:53
if it's insured buffers then it can vacuum it and it's also that's a cash
2:58:56
vacuum it and it's also that's a cash
2:58:56
vacuum it and it's also that's a cash hit right so we didn't have to actually
2:58:58
hit right so we didn't have to actually
2:58:58
hit right so we didn't have to actually do IO
2:59:01
do IO
2:59:01
do IO um now this is what's this is where it
2:59:03
um now this is what's this is where it
2:59:03
um now this is what's this is where it gets different from the regular update
2:59:05
gets different from the regular update
2:59:05
gets different from the regular update insert workflow we were talking about so
2:59:07
insert workflow we were talking about so
2:59:07
insert workflow we were talking about so if we need to actually do i o so we need
2:59:10
if we need to actually do i o so we need
2:59:10
if we need to actually do i o so we need to go out and either and read in the
2:59:14
to go out and either and read in the
2:59:14
to go out and either and read in the block that we're going to vacuum we need
2:59:15
block that we're going to vacuum we need
2:59:15
block that we're going to vacuum we need to find a shared buffer for it that's
2:59:17
to find a shared buffer for it that's
2:59:17
to find a shared buffer for it that's the same as before but the difference
2:59:18
the same as before but the difference
2:59:18
the same as before but the difference here is that vacuum is going to use a
2:59:21
here is that vacuum is going to use a
2:59:21
here is that vacuum is going to use a buffer access strategy and what that
2:59:23
buffer access strategy and what that
2:59:23
buffer access strategy and what that means is that we are going to cap the
2:59:26
means is that we are going to cap the
2:59:26
means is that we are going to cap the number of or so we're going to use a
2:59:28
number of or so we're going to use a
2:59:28
number of or so we're going to use a ring buffer with a certain number of
2:59:30
ring buffer with a certain number of
2:59:30
ring buffer with a certain number of buffers in it and each time we need a
2:59:33
buffers in it and each time we need a
2:59:33
buffers in it and each time we need a new buffer we're gonna go around that
2:59:35
new buffer we're gonna go around that
2:59:35
new buffer we're gonna go around that ring and reuse buffers we've already
2:59:37
ring and reuse buffers we've already
2:59:37
ring and reuse buffers we've already used to vacuum previous blocks and what
2:59:40
used to vacuum previous blocks and what
2:59:40
used to vacuum previous blocks and what this does is it keeps us from using up
2:59:42
this does is it keeps us from using up
2:59:42
this does is it keeps us from using up all of shared buffers just to vacuum
2:59:44
all of shared buffers just to vacuum
2:59:44
all of shared buffers just to vacuum because that that'll you know wash out
2:59:46
because that that'll you know wash out
2:59:46
because that that'll you know wash out all of our working set from share
2:59:48
all of our working set from share
2:59:48
all of our working set from share buffers we don't want that and so the
2:59:50
buffers we don't want that and so the
2:59:50
buffers we don't want that and so the important thing is that when we're
2:59:52
important thing is that when we're
2:59:52
important thing is that when we're initially finding buffers to use we do
2:59:56
initially finding buffers to use we do
2:59:56
initially finding buffers to use we do this lazily on demand so we get a shared
2:59:59
this lazily on demand so we get a shared
2:59:59
this lazily on demand so we get a shared buffer we evict it which counts as an
3:00:01
buffer we evict it which counts as an
3:00:01
buffer we evict it which counts as an eviction in PG statio and now we add it
3:00:04
eviction in PG statio and now we add it
3:00:04
eviction in PG statio and now we add it to the ring
3:00:06
to the ring
3:00:06
to the ring once we've filled up the ring the next
3:00:09
once we've filled up the ring the next
3:00:09
once we've filled up the ring the next time we need a shared buffer we're going
3:00:11
time we need a shared buffer we're going
3:00:11
time we need a shared buffer we're going to reuse the a buffer that we've already
3:00:15
to reuse the a buffer that we've already
3:00:15
to reuse the a buffer that we've already used and that counts as a reuse in PG
3:00:18
used and that counts as a reuse in PG
3:00:18
used and that counts as a reuse in PG statio and when we reuse that buffer we
3:00:23
statio and when we reuse that buffer we
3:00:23
statio and when we reuse that buffer we need to flush the dirty data that was
3:00:25
need to flush the dirty data that was
3:00:25
need to flush the dirty data that was there already and we need to write out
3:00:27
there already and we need to write out
3:00:27
there already and we need to write out the associated wall up to the Allison
3:00:30
the associated wall up to the Allison
3:00:30
the associated wall up to the Allison that was associated with that data that
3:00:34
that was associated with that data that
3:00:34
that was associated with that data that we vacuumed so
3:00:35
we vacuumed so
3:00:35
we vacuumed so what's important here is that even if
3:00:38
what's important here is that even if
3:00:38
what's important here is that even if there are clean shared buffers we're not
3:00:40
there are clean shared buffers we're not
3:00:40
there are clean shared buffers we're not going to go and use them for that data
3:00:41
going to go and use them for that data
3:00:41
going to go and use them for that data we're vacuuming because we're using this
3:00:44
we're vacuuming because we're using this
3:00:44
we're vacuuming because we're using this buffer access strategy
3:00:46
buffer access strategy
3:00:46
buffer access strategy and then what we'll do is you know read
3:00:49
and then what we'll do is you know read
3:00:49
and then what we'll do is you know read the
3:00:50
the
3:00:50
the block that we're trying to vacuum to
3:00:52
block that we're trying to vacuum to
3:00:52
block that we're trying to vacuum to begin with into the buffer that we've
3:00:54
begin with into the buffer that we've
3:00:54
begin with into the buffer that we've selected vacuum it and mark the buffer
3:00:57
selected vacuum it and mark the buffer
3:00:57
selected vacuum it and mark the buffer dirty
3:00:59
dirty
3:00:59
dirty so what does this mean about ypg statio
3:01:03
so what does this mean about ypg statio
3:01:03
so what does this mean about ypg statio is useful so
3:01:06
is useful so
3:01:06
is useful so Auto vacuum is uh their Auto vacuum is
3:01:11
Auto vacuum is uh their Auto vacuum is
3:01:11
Auto vacuum is uh their Auto vacuum is done by Auto vacuum workers so Auto
3:01:13
done by Auto vacuum workers so Auto
3:01:13
done by Auto vacuum workers so Auto vacuum worker is a different back-end
3:01:15
vacuum worker is a different back-end
3:01:15
vacuum worker is a different back-end type than client a client backend so
3:01:19
type than client a client backend so
3:01:19
type than client a client backend so we're able to see that I O pattern and
3:01:21
we're able to see that I O pattern and
3:01:21
we're able to see that I O pattern and that I O that's being done by Auto
3:01:24
that I O that's being done by Auto
3:01:24
that I O that's being done by Auto vacuum separately in PG statio and what
3:01:27
vacuum separately in PG statio and what
3:01:27
vacuum separately in PG statio and what you saw was that we're using this buffer
3:01:29
you saw was that we're using this buffer
3:01:29
you saw was that we're using this buffer access strategy which is separate a
3:01:32
access strategy which is separate a
3:01:32
access strategy which is separate a separate pattern and going to show up
3:01:34
separate pattern and going to show up
3:01:34
separate pattern and going to show up separately then normal transactional
3:01:37
separately then normal transactional
3:01:37
separately then normal transactional workload working set and we may be
3:01:41
workload working set and we may be
3:01:41
workload working set and we may be vacuuming relations that are not part of
3:01:44
vacuuming relations that are not part of
3:01:44
vacuuming relations that are not part of our working set it's not the hottest
3:01:45
our working set it's not the hottest
3:01:45
our working set it's not the hottest data and that's okay so we don't
3:01:47
data and that's okay so we don't
3:01:47
data and that's okay so we don't necessarily want to tune our database to
3:01:50
necessarily want to tune our database to
3:01:50
necessarily want to tune our database to accommodate what we need to vacuum so
3:01:53
accommodate what we need to vacuum so
3:01:53
accommodate what we need to vacuum so it's okay to you know vacuum older data
3:01:55
it's okay to you know vacuum older data
3:01:55
it's okay to you know vacuum older data we don't necessarily need to resize
3:01:57
we don't necessarily need to resize
3:01:57
we don't necessarily need to resize shared buffers so that that all fits in
3:01:59
shared buffers so that that all fits in
3:01:59
shared buffers so that that all fits in shared buffers
3:02:01
shared buffers
3:02:01
shared buffers and similarly context matters so vacuum
3:02:06
and similarly context matters so vacuum
3:02:06
and similarly context matters so vacuum is also a i o context so we can kind of
3:02:09
is also a i o context so we can kind of
3:02:09
is also a i o context so we can kind of see that separated out but another
3:02:12
see that separated out but another
3:02:12
see that separated out but another example of that is a bulk read so for
3:02:15
example of that is a bulk read so for
3:02:15
example of that is a bulk read so for example if you do a large select which
3:02:17
example if you do a large select which
3:02:17
example if you do a large select which is basically a relation with that where
3:02:19
is basically a relation with that where
3:02:19
is basically a relation with that where the number of blocks is
3:02:21
the number of blocks is
3:02:21
the number of blocks is uh you know greater than share buffers
3:02:24
uh you know greater than share buffers
3:02:24
uh you know greater than share buffers divided by four so if we're doing a
3:02:25
divided by four so if we're doing a
3:02:25
divided by four so if we're doing a large select
3:02:27
large select
3:02:27
large select we're going to be doing a lot of reads
3:02:30
we're going to be doing a lot of reads
3:02:30
we're going to be doing a lot of reads and that data is not necessarily part of
3:02:33
and that data is not necessarily part of
3:02:33
and that data is not necessarily part of our working set so we don't necessarily
3:02:35
our working set so we don't necessarily
3:02:35
our working set so we don't necessarily want to resize shared buffers so that
3:02:38
want to resize shared buffers so that
3:02:38
want to resize shared buffers so that that giant table can fit it you might
3:02:40
that giant table can fit it you might
3:02:40
that giant table can fit it you might have a mixed workload with transactional
3:02:41
have a mixed workload with transactional
3:02:41
have a mixed workload with transactional queries alongside some type of
3:02:44
queries alongside some type of
3:02:44
queries alongside some type of analytical work and so what we're trying
3:02:47
analytical work and so what we're trying
3:02:47
analytical work and so what we're trying to do is let's tune specifically for our
3:02:49
to do is let's tune specifically for our
3:02:49
to do is let's tune specifically for our transactional workload for that um that
3:02:53
transactional workload for that um that
3:02:53
transactional workload for that um that working set to fit in memory and to make
3:02:54
working set to fit in memory and to make
3:02:54
working set to fit in memory and to make that workload efficient and then you can
3:02:58
that workload efficient and then you can
3:02:58
that workload efficient and then you can separately think about tuning for your
3:02:59
separately think about tuning for your
3:03:00
separately think about tuning for your analytical workload and make different
3:03:01
analytical workload and make different
3:03:01
analytical workload and make different considerations if you know that a large
3:03:04
considerations if you know that a large
3:03:04
considerations if you know that a large part of your workload is not going to
3:03:05
part of your workload is not going to
3:03:05
part of your workload is not going to fit in shared buffers you might want to
3:03:08
fit in shared buffers you might want to
3:03:08
fit in shared buffers you might want to actually decrease sharp buffers because
3:03:10
actually decrease sharp buffers because
3:03:10
actually decrease sharp buffers because there are certain advantages to that and
3:03:12
there are certain advantages to that and
3:03:12
there are certain advantages to that and I won't go into that more now but
3:03:16
I won't go into that more now but
3:03:16
I won't go into that more now but now given what I just talked about I
3:03:18
now given what I just talked about I
3:03:18
now given what I just talked about I think it's useful to kind of take it
3:03:20
think it's useful to kind of take it
3:03:20
think it's useful to kind of take it back to PG statio and look at a few
3:03:24
back to PG statio and look at a few
3:03:24
back to PG statio and look at a few concrete examples and how it looks in
3:03:26
concrete examples and how it looks in
3:03:26
concrete examples and how it looks in The View
3:03:29
so
3:03:31
so
3:03:31
so I just mentioned that sometimes your
3:03:34
I just mentioned that sometimes your
3:03:34
I just mentioned that sometimes your workload is not going to fit in shared
3:03:37
workload is not going to fit in shared
3:03:37
workload is not going to fit in shared buffers like that's just a reality might
3:03:39
buffers like that's just a reality might
3:03:39
buffers like that's just a reality might not even you know might not fit in
3:03:41
not even you know might not fit in
3:03:41
not even you know might not fit in memory you have a large
3:03:43
memory you have a large
3:03:43
memory you have a large um you have a large database that's like
3:03:45
um you have a large database that's like
3:03:45
um you have a large database that's like pretty normal so in that case it might
3:03:47
pretty normal so in that case it might
3:03:47
pretty normal so in that case it might not be an option for you to increase
3:03:49
not be an option for you to increase
3:03:49
not be an option for you to increase share buffers so that everything fits in
3:03:52
share buffers so that everything fits in
3:03:52
share buffers so that everything fits in shared buffers but one thing that you do
3:03:55
shared buffers but one thing that you do
3:03:55
shared buffers but one thing that you do want to watch out for is that you still
3:03:58
want to watch out for is that you still
3:03:58
want to watch out for is that you still want to avoid client back-ends doing uh
3:04:01
want to avoid client back-ends doing uh
3:04:01
want to avoid client back-ends doing uh their own rights as much as possible so
3:04:05
their own rights as much as possible so
3:04:05
their own rights as much as possible so in this example you can see client back
3:04:07
in this example you can see client back
3:04:07
in this example you can see client back end normal relation rights is decently
3:04:12
end normal relation rights is decently
3:04:12
end normal relation rights is decently high and we actually want that number to
3:04:14
high and we actually want that number to
3:04:15
high and we actually want that number to basically be zero because this is not a
3:04:18
basically be zero because this is not a
3:04:18
basically be zero because this is not a right in the sort of the sense of insert
3:04:21
right in the sort of the sense of insert
3:04:21
right in the sort of the sense of insert update whatever this is actually the
3:04:24
update whatever this is actually the
3:04:24
update whatever this is actually the client looking for a buffer it can't
3:04:27
client looking for a buffer it can't
3:04:27
client looking for a buffer it can't find a clean buffer it has to flush the
3:04:29
find a clean buffer it has to flush the
3:04:29
find a clean buffer it has to flush the data that's in that buffer so we want
3:04:31
data that's in that buffer so we want
3:04:31
data that's in that buffer so we want client back ends to be doing zero rights
3:04:34
client back ends to be doing zero rights
3:04:34
client back ends to be doing zero rights we want the checkpointer and the
3:04:37
we want the checkpointer and the
3:04:37
we want the checkpointer and the background writer to be taking care of
3:04:38
background writer to be taking care of
3:04:38
background writer to be taking care of this for them so that when you do a ever
3:04:41
this for them so that when you do a ever
3:04:41
this for them so that when you do a ever a read you're not doing it right first
3:04:44
a read you're not doing it right first
3:04:44
a read you're not doing it right first um so in this example we're seeing a
3:04:48
um so in this example we're seeing a
3:04:48
um so in this example we're seeing a background writer is doing some rights
3:04:50
background writer is doing some rights
3:04:50
background writer is doing some rights but maybe it could be doing more so we
3:04:52
but maybe it could be doing more so we
3:04:52
but maybe it could be doing more so we have some options there
3:04:54
have some options there
3:04:54
have some options there um background writer historically has a
3:04:56
um background writer historically has a
3:04:56
um background writer historically has a certain throughput cap that's relatively
3:04:58
certain throughput cap that's relatively
3:04:58
certain throughput cap that's relatively low so if you have a super you know fast
3:05:02
low so if you have a super you know fast
3:05:02
low so if you have a super you know fast like high throughput workload then maybe
3:05:06
like high throughput workload then maybe
3:05:06
like high throughput workload then maybe background writer is not going to be
3:05:07
background writer is not going to be
3:05:07
background writer is not going to be able to help you but
3:05:09
able to help you but
3:05:09
able to help you but um you can decrease the background
3:05:11
um you can decrease the background
3:05:11
um you can decrease the background writer delay you can increase background
3:05:13
writer delay you can increase background
3:05:13
writer delay you can increase background writer lru Max pages I won't get into
3:05:15
writer lru Max pages I won't get into
3:05:15
writer lru Max pages I won't get into the specific mechanics of those uh gucks
3:05:18
the specific mechanics of those uh gucks
3:05:18
the specific mechanics of those uh gucks but there's articles online you can look
3:05:21
but there's articles online you can look
3:05:21
but there's articles online you can look at background writer tuning that kind of
3:05:22
at background writer tuning that kind of
3:05:22
at background writer tuning that kind of thing so
3:05:25
thing so
3:05:25
thing so um
3:05:25
um
3:05:25
um and then a lot of times though like
3:05:29
and then a lot of times though like
3:05:29
and then a lot of times though like let's take this case shared buffers is
3:05:31
let's take this case shared buffers is
3:05:31
let's take this case shared buffers is too small a lot of people are if you
3:05:34
too small a lot of people are if you
3:05:34
too small a lot of people are if you have the ability to change share buffers
3:05:36
have the ability to change share buffers
3:05:36
have the ability to change share buffers of course you need to restart and
3:05:38
of course you need to restart and
3:05:38
of course you need to restart and everything
3:05:39
everything
3:05:39
everything um a lot of times that's going to help
3:05:41
um a lot of times that's going to help
3:05:41
um a lot of times that's going to help with your i o problem so in this example
3:05:43
with your i o problem so in this example
3:05:43
with your i o problem so in this example in contrast if you look at the number of
3:05:46
in contrast if you look at the number of
3:05:46
in contrast if you look at the number of reads for client back in in the normal
3:05:48
reads for client back in in the normal
3:05:48
reads for client back in in the normal context and the number of Rights
3:05:51
context and the number of Rights
3:05:51
context and the number of Rights and for a client back in a normal
3:05:53
and for a client back in a normal
3:05:53
and for a client back in a normal context there's almost one right for
3:05:56
context there's almost one right for
3:05:56
context there's almost one right for every read so that's bad that's a bad
3:05:59
every read so that's bad that's a bad
3:05:59
every read so that's bad that's a bad sign that means most likely we are
3:06:01
sign that means most likely we are
3:06:01
sign that means most likely we are evicting and re-reading in the same
3:06:03
evicting and re-reading in the same
3:06:03
evicting and re-reading in the same blocks over and over and you see
3:06:05
blocks over and over and you see
3:06:05
blocks over and over and you see evictions as High
3:06:07
evictions as High
3:06:07
evictions as High um you know after the first time that
3:06:09
um you know after the first time that
3:06:09
um you know after the first time that we've gotten everything off the free
3:06:10
we've gotten everything off the free
3:06:10
we've gotten everything off the free list and we've used every buffer at
3:06:13
list and we've used every buffer at
3:06:13
list and we've used every buffer at least once everything is going to be an
3:06:15
least once everything is going to be an
3:06:15
least once everything is going to be an eviction after that so but in this case
3:06:18
eviction after that so but in this case
3:06:18
eviction after that so but in this case because we're seeing reads and writes be
3:06:20
because we're seeing reads and writes be
3:06:20
because we're seeing reads and writes be very proportionate that's a proportional
3:06:23
very proportionate that's a proportional
3:06:23
very proportionate that's a proportional that's like a hint that hey or a signal
3:06:26
that's like a hint that hey or a signal
3:06:27
that's like a hint that hey or a signal that it may be necessary to increase
3:06:29
that it may be necessary to increase
3:06:29
that it may be necessary to increase sharp buffers there are primary workings
3:06:32
sharp buffers there are primary workings
3:06:32
sharp buffers there are primary workings that just isn't fitting and you can see
3:06:35
that just isn't fitting and you can see
3:06:35
that just isn't fitting and you can see that with our cachet ratio too it's
3:06:36
that with our cachet ratio too it's
3:06:36
that with our cachet ratio too it's around 60 when we look at the client
3:06:39
around 60 when we look at the client
3:06:39
around 60 when we look at the client back in normal hits and reads
3:06:42
back in normal hits and reads
3:06:42
back in normal hits and reads and just a simple cachet ratio query for
3:06:47
and just a simple cachet ratio query for
3:06:47
and just a simple cachet ratio query for client Norm backends uh doing things in
3:06:50
client Norm backends uh doing things in
3:06:50
client Norm backends uh doing things in the normal context with permanent
3:06:52
the normal context with permanent
3:06:52
the normal context with permanent relations you can calculate the cachet
3:06:54
relations you can calculate the cachet
3:06:54
relations you can calculate the cachet ratio this way
3:06:56
ratio this way
3:06:56
ratio this way on the other hand you want to avoid
3:06:58
on the other hand you want to avoid
3:06:58
on the other hand you want to avoid premature optimization
3:07:01
premature optimization
3:07:01
premature optimization so one of the things that was really
3:07:03
so one of the things that was really
3:07:03
so one of the things that was really hard before PG statio was actually
3:07:06
hard before PG statio was actually
3:07:06
hard before PG statio was actually calculating your cachet ratio if you
3:07:08
calculating your cachet ratio if you
3:07:08
calculating your cachet ratio if you have any sort of other types of i o then
3:07:10
have any sort of other types of i o then
3:07:10
have any sort of other types of i o then you're very standard uh transactional
3:07:13
you're very standard uh transactional
3:07:13
you're very standard uh transactional workload going on so in this example you
3:07:16
workload going on so in this example you
3:07:16
workload going on so in this example you can see that client back-end normal
3:07:18
can see that client back-end normal
3:07:18
can see that client back-end normal context reads pretty low looks like we
3:07:22
context reads pretty low looks like we
3:07:22
context reads pretty low looks like we read in our working set and then the
3:07:24
read in our working set and then the
3:07:24
read in our working set and then the hits are really high we just keep
3:07:25
hits are really high we just keep
3:07:25
hits are really high we just keep reusing those
3:07:28
reusing those
3:07:28
reusing those um those blocks and you know everything
3:07:31
um those blocks and you know everything
3:07:31
um those blocks and you know everything seems like it's kind of going okay in
3:07:34
seems like it's kind of going okay in
3:07:34
seems like it's kind of going okay in terms of that but you can tell that the
3:07:36
terms of that but you can tell that the
3:07:36
terms of that but you can tell that the number of client backend bulk reads is
3:07:39
number of client backend bulk reads is
3:07:39
number of client backend bulk reads is really high we also have a fair number
3:07:40
really high we also have a fair number
3:07:40
really high we also have a fair number of Auto vacuum reads so what you can see
3:07:43
of Auto vacuum reads so what you can see
3:07:43
of Auto vacuum reads so what you can see there is if those were all together they
3:07:46
there is if those were all together they
3:07:46
there is if those were all together they weren't separated by back end type and
3:07:48
weren't separated by back end type and
3:07:48
weren't separated by back end type and context then we would actually calculate
3:07:52
context then we would actually calculate
3:07:52
context then we would actually calculate a pretty incorrect cachet ratio and I'll
3:07:54
a pretty incorrect cachet ratio and I'll
3:07:54
a pretty incorrect cachet ratio and I'll show you what that would look like
3:07:57
show you what that would look like
3:07:57
show you what that would look like if we use PG stats database to calculate
3:08:01
if we use PG stats database to calculate
3:08:01
if we use PG stats database to calculate the cachet ratio here we get about 45
3:08:04
the cachet ratio here we get about 45
3:08:04
the cachet ratio here we get about 45 and that's similar to if you use PG
3:08:07
and that's similar to if you use PG
3:08:07
and that's similar to if you use PG statio without a where Clause but what
3:08:09
statio without a where Clause but what
3:08:09
statio without a where Clause but what we care about is tuning for our regular
3:08:13
we care about is tuning for our regular
3:08:13
we care about is tuning for our regular workload so once we add in the where
3:08:16
workload so once we add in the where
3:08:16
workload so once we add in the where Clause where the client is back with the
3:08:18
Clause where the client is back with the
3:08:18
Clause where the client is back with the backend type is a client backend it's in
3:08:20
backend type is a client backend it's in
3:08:20
backend type is a client backend it's in the normal context it's a permanent no
3:08:24
the normal context it's a permanent no
3:08:24
the normal context it's a permanent no temp relations now caches ratio is
3:08:26
temp relations now caches ratio is
3:08:26
temp relations now caches ratio is basically 99 I mean we definitely don't
3:08:28
basically 99 I mean we definitely don't
3:08:28
basically 99 I mean we definitely don't need to increase share buffers for this
3:08:30
need to increase share buffers for this
3:08:30
need to increase share buffers for this so I think pts.io gives you the
3:08:34
so I think pts.io gives you the
3:08:34
so I think pts.io gives you the additional information that you need to
3:08:37
additional information that you need to
3:08:37
additional information that you need to avoid these kinds of premature
3:08:39
avoid these kinds of premature
3:08:39
avoid these kinds of premature optimizations
3:08:42
so what's next to refugees.io um so
3:08:46
so what's next to refugees.io um so
3:08:46
so what's next to refugees.io um so hopefully in postgres 17 we're going to
3:08:48
hopefully in postgres 17 we're going to
3:08:48
hopefully in postgres 17 we're going to add there's a category of missing i o
3:08:51
add there's a category of missing i o
3:08:51
add there's a category of missing i o that I think
3:08:53
that I think
3:08:53
that I think um would be really good to add which is
3:08:54
um would be really good to add which is
3:08:54
um would be really good to add which is all the i o that's outside of uh shared
3:08:57
all the i o that's outside of uh shared
3:08:57
all the i o that's outside of uh shared buffers uh and outside of local buffers
3:09:01
buffers uh and outside of local buffers
3:09:01
buffers uh and outside of local buffers so things like if you do you know
3:09:02
so things like if you do you know
3:09:03
so things like if you do you know there's certain operations like create
3:09:04
there's certain operations like create
3:09:04
there's certain operations like create index where it's doing a lot of i o sort
3:09:07
index where it's doing a lot of i o sort
3:09:07
index where it's doing a lot of i o sort of directly
3:09:09
of directly
3:09:09
of directly um we're calling it bypass i o also per
3:09:11
um we're calling it bypass i o also per
3:09:11
um we're calling it bypass i o also per connection i o stats so taking PG statio
3:09:13
connection i o stats so taking PG statio
3:09:13
connection i o stats so taking PG statio separating out per connection and then
3:09:16
separating out per connection and then
3:09:16
separating out per connection and then integrating PG sat wall the information
3:09:18
integrating PG sat wall the information
3:09:18
integrating PG sat wall the information that's there into PG statio and sort of
3:09:21
that's there into PG statio and sort of
3:09:21
that's there into PG statio and sort of streamlining it so there's a single
3:09:22
streamlining it so there's a single
3:09:22
streamlining it so there's a single source of Truth for your i o information
3:09:26
source of Truth for your i o information
3:09:26
source of Truth for your i o information um I will be answering questions on
3:09:28
um I will be answering questions on
3:09:28
um I will be answering questions on Discord so I'd love to hear about your
3:09:31
Discord so I'd love to hear about your
3:09:31
Discord so I'd love to hear about your use cases your questions and I'm just
3:09:33
use cases your questions and I'm just
3:09:33
use cases your questions and I'm just really excited for people to use the
3:09:35
really excited for people to use the
3:09:35
really excited for people to use the view in postgres 16 and tell me what
3:09:37
view in postgres 16 and tell me what
3:09:37
view in postgres 16 and tell me what they think
3:09:42
that was wonderful
3:09:45
that was wonderful
3:09:45
that was wonderful thank you Melanie for being here
3:09:48
thank you Melanie for being here
3:09:48
thank you Melanie for being here um I have a question for you for someone
3:09:49
um I have a question for you for someone
3:09:49
um I have a question for you for someone who's really familiar with PG stat
3:09:51
who's really familiar with PG stat
3:09:51
who's really familiar with PG stat statements
3:09:53
statements
3:09:53
statements um when should they think about using PG
3:09:55
um when should they think about using PG
3:09:55
um when should they think about using PG stat IO versus PG stat statements right
3:09:58
stat IO versus PG stat statements right
3:09:58
stat IO versus PG stat statements right so if you do stat statements is still
3:10:00
so if you do stat statements is still
3:10:00
so if you do stat statements is still going to be your go-to when you care
3:10:03
going to be your go-to when you care
3:10:03
going to be your go-to when you care about per statement per query
3:10:05
about per statement per query
3:10:05
about per statement per query information so if you kind of already
3:10:07
information so if you kind of already
3:10:07
information so if you kind of already know like this query is really slow then
3:10:10
know like this query is really slow then
3:10:10
know like this query is really slow then you want to use PG stat statements for
3:10:12
you want to use PG stat statements for
3:10:12
you want to use PG stat statements for sure
3:10:13
sure
3:10:13
sure okay so the other thing I think people
3:10:15
okay so the other thing I think people
3:10:15
okay so the other thing I think people always wonder about with uh you know I O
3:10:18
always wonder about with uh you know I O
3:10:18
always wonder about with uh you know I O related tooling and postgres and
3:10:20
related tooling and postgres and
3:10:20
related tooling and postgres and specifically in this case
3:10:22
specifically in this case
3:10:22
specifically in this case so once this is out there like
3:10:24
so once this is out there like
3:10:24
so once this is out there like presumably I'll see something if I'm in
3:10:25
presumably I'll see something if I'm in
3:10:25
presumably I'll see something if I'm in a cloud provider like I'll be able to
3:10:27
a cloud provider like I'll be able to
3:10:27
a cloud provider like I'll be able to actually get some data because the
3:10:28
actually get some data because the
3:10:28
actually get some data because the underlying stuff is is often hard to
3:10:30
underlying stuff is is often hard to
3:10:30
underlying stuff is is often hard to understand right yeah so that's the
3:10:32
understand right yeah so that's the
3:10:32
understand right yeah so that's the great thing about PG studio is that um
3:10:35
great thing about PG studio is that um
3:10:35
great thing about PG studio is that um it will on any cloud provider unless
3:10:37
it will on any cloud provider unless
3:10:37
it will on any cloud provider unless they restrict it for some reason you'll
3:10:39
they restrict it for some reason you'll
3:10:39
they restrict it for some reason you'll be able to see it uh we have put all the
3:10:42
be able to see it uh we have put all the
3:10:42
be able to see it uh we have put all the collection points in a place where it
3:10:45
collection points in a place where it
3:10:45
collection points in a place where it doesn't matter what the underlying
3:10:47
doesn't matter what the underlying
3:10:47
doesn't matter what the underlying storage type is or anything like that so
3:10:49
storage type is or anything like that so
3:10:49
storage type is or anything like that so it'll just work on any cloud provider
3:10:51
it'll just work on any cloud provider
3:10:51
it'll just work on any cloud provider awesome
3:10:53
awesome
3:10:53
awesome so I know you just um committed PG stat
3:10:56
so I know you just um committed PG stat
3:10:56
so I know you just um committed PG stat i o to postgres 16. I'm just curious
3:10:59
i o to postgres 16. I'm just curious
3:10:59
i o to postgres 16. I'm just curious what's next for you and postgres are you
3:11:01
what's next for you and postgres are you
3:11:01
what's next for you and postgres are you going to continue to work in
3:11:02
going to continue to work in
3:11:02
going to continue to work in observability or I think that one of the
3:11:05
observability or I think that one of the
3:11:05
observability or I think that one of the things that I'm really excited about is
3:11:07
things that I'm really excited about is
3:11:07
things that I'm really excited about is postgres 15 actually added change the
3:11:10
postgres 15 actually added change the
3:11:10
postgres 15 actually added change the underlying statistics system to use
3:11:13
underlying statistics system to use
3:11:13
underlying statistics system to use shared memory and the way that it is now
3:11:16
shared memory and the way that it is now
3:11:16
shared memory and the way that it is now it's much more reliable and um the
3:11:20
it's much more reliable and um the
3:11:20
it's much more reliable and um the adding new statistics is a lot easier
3:11:22
adding new statistics is a lot easier
3:11:22
adding new statistics is a lot easier and sort of straightforward so what I
3:11:24
and sort of straightforward so what I
3:11:24
and sort of straightforward so what I think what I'd really love to see is
3:11:26
think what I'd really love to see is
3:11:26
think what I'd really love to see is users who are not you know necessarily
3:11:28
users who are not you know necessarily
3:11:28
users who are not you know necessarily full-time postgres contributors but know
3:11:30
full-time postgres contributors but know
3:11:30
full-time postgres contributors but know what their use cases are
3:11:32
what their use cases are
3:11:32
what their use cases are so for them to add new statistics and so
3:11:35
so for them to add new statistics and so
3:11:35
so for them to add new statistics and so like I'm excited to review more patches
3:11:37
like I'm excited to review more patches
3:11:37
like I'm excited to review more patches around observability
3:11:39
around observability
3:11:39
around observability um I know some of the folks at dalibur
3:11:41
um I know some of the folks at dalibur
3:11:41
um I know some of the folks at dalibur proposed some uh exciting news
3:11:43
proposed some uh exciting news
3:11:43
proposed some uh exciting news statistics around parallel query for
3:11:45
statistics around parallel query for
3:11:45
statistics around parallel query for postgres 17
3:11:47
postgres 17
3:11:47
postgres 17 um below one of my co-workers here who
3:11:49
um below one of my co-workers here who
3:11:49
um below one of my co-workers here who did a poseiduscon talk on CI he's
3:11:52
did a poseiduscon talk on CI he's
3:11:52
did a poseiduscon talk on CI he's thinking about working on some of the
3:11:54
thinking about working on some of the
3:11:54
thinking about working on some of the items that I the open items that I
3:11:56
items that I the open items that I
3:11:56
items that I the open items that I listed
3:11:57
listed
3:11:57
listed um so I think like the the goal is to
3:12:00
um so I think like the the goal is to
3:12:00
um so I think like the the goal is to have you know people that are using
3:12:03
have you know people that are using
3:12:03
have you know people that are using postgresp being like every single day
3:12:05
postgresp being like every single day
3:12:05
postgresp being like every single day you know in in the trenches adding the
3:12:08
you know in in the trenches adding the
3:12:08
you know in in the trenches adding the new observability
3:12:10
new observability
3:12:10
new observability um statistics awesome and people can
3:12:14
um statistics awesome and people can
3:12:14
um statistics awesome and people can reach you obviously today during the
3:12:16
reach you obviously today during the
3:12:16
reach you obviously today during the event on Discord if they have questions
3:12:17
event on Discord if they have questions
3:12:17
event on Discord if they have questions and then more generally if someone
3:12:19
and then more generally if someone
3:12:19
and then more generally if someone watches this talk on YouTube in three
3:12:21
watches this talk on YouTube in three
3:12:21
watches this talk on YouTube in three months and they have an idea for some
3:12:23
months and they have an idea for some
3:12:23
months and they have an idea for some statistics or feedback on observability
3:12:25
statistics or feedback on observability
3:12:25
statistics or feedback on observability how do they reach you well you can tweet
3:12:28
how do they reach you well you can tweet
3:12:28
how do they reach you well you can tweet at me I'm not like that prolific so I
3:12:31
at me I'm not like that prolific so I
3:12:31
at me I'm not like that prolific so I will probably answer but I think the
3:12:33
will probably answer but I think the
3:12:33
will probably answer but I think the best thing to do would be on the hackers
3:12:35
best thing to do would be on the hackers
3:12:35
best thing to do would be on the hackers mailing list you can either email me
3:12:37
mailing list you can either email me
3:12:37
mailing list you can either email me directly or you can send an email to the
3:12:40
directly or you can send an email to the
3:12:40
directly or you can send an email to the Hacker's mailing list and propose your
3:12:42
Hacker's mailing list and propose your
3:12:42
Hacker's mailing list and propose your idea and sort of don't be afraid to like
3:12:45
idea and sort of don't be afraid to like
3:12:45
idea and sort of don't be afraid to like but I like don't be afraid to email the
3:12:47
but I like don't be afraid to email the
3:12:47
but I like don't be afraid to email the hackers mailing list because you have a
3:12:48
hackers mailing list because you have a
3:12:48
hackers mailing list because you have a higher chance that people respond that
3:12:50
higher chance that people respond that
3:12:50
higher chance that people respond that if you want help like composing your
3:12:52
if you want help like composing your
3:12:52
if you want help like composing your email because I know it could be
3:12:53
email because I know it could be
3:12:53
email because I know it could be intimidating I'm happy to help people
3:12:56
intimidating I'm happy to help people
3:12:56
intimidating I'm happy to help people with formatting and wording
3:13:00
with formatting and wording
3:13:00
with formatting and wording awesome well I'm so glad
3:13:03
awesome well I'm so glad
3:13:03
awesome well I'm so glad um that you had this time on your
3:13:05
um that you had this time on your
3:13:05
um that you had this time on your schedule and the ability to come and
3:13:07
schedule and the ability to come and
3:13:07
schedule and the ability to come and give the talk today thank you very much
3:13:08
give the talk today thank you very much
3:13:08
give the talk today thank you very much Melanie thank you
3:13:10
Melanie thank you
3:13:10
Melanie thank you I also wanted to say that people who
3:13:13
I also wanted to say that people who
3:13:13
I also wanted to say that people who have feedback on your talk or any of the
3:13:15
have feedback on your talk or any of the
3:13:15
have feedback on your talk or any of the other talks we've been popping a banner
3:13:16
other talks we've been popping a banner
3:13:16
other talks we've been popping a banner up with the aka.ms URL to fill out the
3:13:19
up with the aka.ms URL to fill out the
3:13:19
up with the aka.ms URL to fill out the attendee survey and you can fill it out
3:13:21
attendee survey and you can fill it out
3:13:21
attendee survey and you can fill it out multiple times so you know on the talks
3:13:23
multiple times so you know on the talks
3:13:23
multiple times so you know on the talks you've seen in today's live stream
3:13:25
you've seen in today's live stream
3:13:25
you've seen in today's live stream tomorrow is the media live stream the
3:13:27
tomorrow is the media live stream the
3:13:27
tomorrow is the media live stream the on-demand talks that it's open until
3:13:29
on-demand talks that it's open until
3:13:29
on-demand talks that it's open until next Friday the 28th or something cool
3:13:33
next Friday the 28th or something cool
3:13:33
next Friday the 28th or something cool all right so um I've also been popping
3:13:36
all right so um I've also been popping
3:13:36
all right so um I've also been popping stickers and swag bag codes in the
3:13:40
stickers and swag bag codes in the
3:13:40
stickers and swag bag codes in the banners during the live stream so those
3:13:42
banners during the live stream so those
3:13:42
banners during the live stream so those of you who care about Swag
3:13:45
of you who care about Swag
3:13:45
of you who care about Swag um definitely pay attention to those
3:13:46
um definitely pay attention to those
3:13:46
um definitely pay attention to those banners as they pop across on each talk
3:13:48
banners as they pop across on each talk
3:13:48
banners as they pop across on each talk so you can make sure to try to get your
3:13:50
so you can make sure to try to get your
3:13:50
so you can make sure to try to get your swag bag or your sticker pack I I think
3:13:53
swag bag or your sticker pack I I think
3:13:53
swag bag or your sticker pack I I think it's time for our sixth speaker today I
3:13:56
it's time for our sixth speaker today I
3:13:56
it's time for our sixth speaker today I believe it is hi
3:13:59
believe it is hi
3:13:59
believe it is hi welcome
3:14:00
welcome
3:14:00
welcome um today
3:14:01
um today
3:14:01
um today yes correctly absolutely for those uh
3:14:06
yes correctly absolutely for those uh
3:14:06
yes correctly absolutely for those uh who don't know Andre uh he's actually I
3:14:08
who don't know Andre uh he's actually I
3:14:09
who don't know Andre uh he's actually I guess a long time postgres hacker I
3:14:10
guess a long time postgres hacker I
3:14:10
guess a long time postgres hacker I guess we'll we'll put you in that
3:14:11
guess we'll we'll put you in that
3:14:12
guess we'll we'll put you in that category now
3:14:13
category now
3:14:13
category now um and has worked on a whole bunch of
3:14:15
um and has worked on a whole bunch of
3:14:15
um and has worked on a whole bunch of different parts of the postgres project
3:14:17
different parts of the postgres project
3:14:17
different parts of the postgres project um most recent patch I had a blog post
3:14:19
um most recent patch I had a blog post
3:14:19
um most recent patch I had a blog post on this a couple of years ago like there
3:14:21
on this a couple of years ago like there
3:14:21
on this a couple of years ago like there should be a better way so now there is
3:14:22
should be a better way so now there is
3:14:22
should be a better way so now there is was adding iteration counts for the
3:14:25
was adding iteration counts for the
3:14:25
was adding iteration counts for the watch command in psql which sounds like
3:14:27
watch command in psql which sounds like
3:14:27
watch command in psql which sounds like a small thing but like it's a thing
3:14:29
a small thing but like it's a thing
3:14:29
a small thing but like it's a thing people need to do and there was really
3:14:30
people need to do and there was really
3:14:30
people need to do and there was really no good way to do it so yeah and uh this
3:14:33
no good way to do it so yeah and uh this
3:14:33
no good way to do it so yeah and uh this feature actually we had to do is on
3:14:37
feature actually we had to do is on
3:14:37
feature actually we had to do is on leave session on like discussing which
3:14:41
leave session on like discussing which
3:14:41
leave session on like discussing which fish which features we could add and he
3:14:44
fish which features we could add and he
3:14:44
fish which features we could add and he proposes that let's do this and we uh
3:14:47
proposes that let's do this and we uh
3:14:47
proposes that let's do this and we uh hacked it online
3:14:49
hacked it online
3:14:49
hacked it online and submitted to postgres hackers and it
3:14:52
and submitted to postgres hackers and it
3:14:52
and submitted to postgres hackers and it was committed it was cool awesome a
3:14:55
was committed it was cool awesome a
3:14:55
was committed it was cool awesome a couple other things you've done a bunch
3:14:57
couple other things you've done a bunch
3:14:57
couple other things you've done a bunch of work on indexes over the years and
3:14:59
of work on indexes over the years and
3:14:59
of work on indexes over the years and you are one of the maintainers of YG is
3:15:01
you are one of the maintainers of YG is
3:15:01
you are one of the maintainers of YG is that correct uh yes uh it was started by
3:15:05
that correct uh yes uh it was started by
3:15:05
that correct uh yes uh it was started by uh Dan Farina but I I later joined the
3:15:09
uh Dan Farina but I I later joined the
3:15:09
uh Dan Farina but I I later joined the team and uh
3:15:10
team and uh
3:15:10
team and uh uh organize it
3:15:13
uh organize it
3:15:13
uh organize it all the stuff and moving around so both
3:15:17
all the stuff and moving around so both
3:15:17
all the stuff and moving around so both of those areas deal with compression
3:15:18
of those areas deal with compression
3:15:18
of those areas deal with compression there's a lot of other parts of
3:15:19
there's a lot of other parts of
3:15:19
there's a lot of other parts of compression in postgres and I think
3:15:22
compression in postgres and I think
3:15:22
compression in postgres and I think that's our topic for today so let's uh I
3:15:24
that's our topic for today so let's uh I
3:15:24
that's our topic for today so let's uh I just want to shine into the wall G when
3:15:27
just want to shine into the wall G when
3:15:27
just want to shine into the wall G when it was first created was Dan Farina was
3:15:29
it was first created was Dan Farina was
3:15:29
it was first created was Dan Farina was absolutely the mentor and the advisor
3:15:31
absolutely the mentor and the advisor
3:15:31
absolutely the mentor and the advisor but it was actually a summer intern at
3:15:33
but it was actually a summer intern at
3:15:33
but it was actually a summer intern at studys data named Katie Lee who I
3:15:35
studys data named Katie Lee who I
3:15:35
studys data named Katie Lee who I believe was a UC Berkeley College
3:15:36
believe was a UC Berkeley College
3:15:36
believe was a UC Berkeley College student at the time so
3:15:38
student at the time so
3:15:38
student at the time so um yeah just wanted to throw that shout
3:15:41
um yeah just wanted to throw that shout
3:15:41
um yeah just wanted to throw that shout out to her all right
3:15:43
out to her all right
3:15:43
out to her all right um Andre over to you
3:15:46
um Andre over to you
3:15:46
um Andre over to you uh so hi everyone my name is Andre I'm
3:15:50
uh so hi everyone my name is Andre I'm
3:15:50
uh so hi everyone my name is Andre I'm doing postgres every day this sounds
3:15:53
doing postgres every day this sounds
3:15:53
doing postgres every day this sounds funny yeah and I'm going to continue uh
3:15:57
funny yeah and I'm going to continue uh
3:15:57
funny yeah and I'm going to continue uh Melanie's talk with
3:16:00
Melanie's talk with
3:16:00
Melanie's talk with some more postgres internals and this
3:16:03
some more postgres internals and this
3:16:03
some more postgres internals and this stock will be uh like interesting to
3:16:06
stock will be uh like interesting to
3:16:06
stock will be uh like interesting to dbas and software developers uh to
3:16:10
dbas and software developers uh to
3:16:10
dbas and software developers uh to enable new features that are already in
3:16:13
enable new features that are already in
3:16:13
enable new features that are already in in postgres core but also this talk
3:16:16
in postgres core but also this talk
3:16:16
in postgres core but also this talk could be of interest to of interest to
3:16:19
could be of interest to of interest to
3:16:19
could be of interest to of interest to uh scientists and someone seeking for
3:16:23
uh scientists and someone seeking for
3:16:23
uh scientists and someone seeking for new ways to improve uh postgres uh in a
3:16:28
new ways to improve uh postgres uh in a
3:16:28
new ways to improve uh postgres uh in a like classical algorithmical way with a
3:16:31
like classical algorithmical way with a
3:16:31
like classical algorithmical way with a compression which is like
3:16:33
compression which is like
3:16:33
compression which is like uh kind of a old field of study
3:16:37
uh kind of a old field of study
3:16:37
uh kind of a old field of study so let's do it on compression of
3:16:40
so let's do it on compression of
3:16:40
so let's do it on compression of everything everything that I could
3:16:43
everything everything that I could
3:16:43
everything everything that I could imagine
3:16:45
what is basic idea of compression that
3:16:48
what is basic idea of compression that
3:16:48
what is basic idea of compression that we like have a function that is reducing
3:16:52
we like have a function that is reducing
3:16:52
we like have a function that is reducing uh reducing size of some input vector
3:16:56
uh reducing size of some input vector
3:16:56
uh reducing size of some input vector and output Vector must be smaller than
3:16:59
and output Vector must be smaller than
3:16:59
and output Vector must be smaller than input
3:17:00
input
3:17:00
input but now since just don't work that way
3:17:03
but now since just don't work that way
3:17:03
but now since just don't work that way in fact uh output is uh in like so many
3:17:10
in fact uh output is uh in like so many
3:17:10
in fact uh output is uh in like so many cases is bigger than input and what's
3:17:12
cases is bigger than input and what's
3:17:12
cases is bigger than input and what's the point then
3:17:14
the point then
3:17:14
the point then uh let's see compression as a some way
3:17:17
uh let's see compression as a some way
3:17:17
uh let's see compression as a some way of a
3:17:18
of a
3:17:18
of a like a process of decompression as uh
3:17:22
like a process of decompression as uh
3:17:22
like a process of decompression as uh some way of decoding three uh so we have
3:17:25
some way of decoding three uh so we have
3:17:25
some way of decoding three uh so we have a decision tree uh and uh each decision
3:17:28
a decision tree uh and uh each decision
3:17:28
a decision tree uh and uh each decision is representing one bit of our input
3:17:33
is representing one bit of our input
3:17:33
is representing one bit of our input now of our compressed input that we want
3:17:35
now of our compressed input that we want
3:17:35
now of our compressed input that we want to decompress uh if this decision tree
3:17:39
to decompress uh if this decision tree
3:17:39
to decompress uh if this decision tree is balanced then uh there is
3:17:45
there is a ratio of exactly one the size
3:17:49
there is a ratio of exactly one the size
3:17:49
there is a ratio of exactly one the size of compressed input equals the size of
3:17:52
of compressed input equals the size of
3:17:52
of compressed input equals the size of decompressed input there is no point in
3:17:53
decompressed input there is no point in
3:17:54
decompressed input there is no point in having a balanced decoding tree so we
3:17:56
having a balanced decoding tree so we
3:17:56
having a balanced decoding tree so we want to uh imbalance decodentry so some
3:18:00
want to uh imbalance decodentry so some
3:18:00
want to uh imbalance decodentry so some passes are short but the vast majority
3:18:02
passes are short but the vast majority
3:18:02
passes are short but the vast majority of passes through the decoding tree is
3:18:05
of passes through the decoding tree is
3:18:05
of passes through the decoding tree is longer than uh pass through the balanced
3:18:08
longer than uh pass through the balanced
3:18:08
longer than uh pass through the balanced uh regarding tree and that's the whole
3:18:11
uh regarding tree and that's the whole
3:18:11
uh regarding tree and that's the whole idea of a compression that we have a
3:18:14
idea of a compression that we have a
3:18:14
idea of a compression that we have a trade-off that some frequent data is
3:18:17
trade-off that some frequent data is
3:18:17
trade-off that some frequent data is shorter but the vast majority of rare
3:18:22
shorter but the vast majority of rare
3:18:22
shorter but the vast majority of rare input vectors is much or at least
3:18:25
input vectors is much or at least
3:18:25
input vectors is much or at least slightly longer than uh average input
3:18:30
slightly longer than uh average input
3:18:30
slightly longer than uh average input and that's why we have to do this uh
3:18:34
and that's why we have to do this uh
3:18:34
and that's why we have to do this uh trade-off wisely we cannot just apply
3:18:37
trade-off wisely we cannot just apply
3:18:37
trade-off wisely we cannot just apply compression everywhere
3:18:38
compression everywhere
3:18:39
compression everywhere but the idea is sensible and it works
3:18:42
but the idea is sensible and it works
3:18:42
but the idea is sensible and it works uh so in recent days uh author of
3:18:46
uh so in recent days uh author of
3:18:46
uh so in recent days uh author of another database because
3:18:49
another database because
3:18:49
another database because told on one of conferences that every
3:18:53
told on one of conferences that every
3:18:53
told on one of conferences that every byte that goes through IO to the D to
3:18:56
byte that goes through IO to the D to
3:18:56
byte that goes through IO to the D to the Discord to the network deserves some
3:18:59
the Discord to the network deserves some
3:18:59
the Discord to the network deserves some compression and I largely agree with
3:19:02
compression and I largely agree with
3:19:02
compression and I largely agree with Alexa
3:19:03
Alexa
3:19:03
Alexa and uh let's talk about compression
3:19:07
and uh let's talk about compression
3:19:07
and uh let's talk about compression deeper uh what do we have in postgres
3:19:09
deeper uh what do we have in postgres
3:19:09
deeper uh what do we have in postgres for compressing stuff
3:19:11
for compressing stuff
3:19:11
for compressing stuff let's talk about codex so coder and
3:19:14
let's talk about codex so coder and
3:19:14
let's talk about codex so coder and coders and decoders we always have these
3:19:17
coders and decoders we always have these
3:19:17
coders and decoders we always have these two functions uh to take a
3:19:20
two functions uh to take a
3:19:20
two functions uh to take a bit of vector of bits and make another
3:19:24
bit of vector of bits and make another
3:19:24
bit of vector of bits and make another Vector of bits
3:19:26
Vector of bits
3:19:26
Vector of bits and vice versa
3:19:28
and vice versa
3:19:29
and vice versa uh usually we use so-called Pareto
3:19:34
uh usually we use so-called Pareto
3:19:34
uh usually we use so-called Pareto Frontier to describe useful codecs
3:19:37
Frontier to describe useful codecs
3:19:37
Frontier to describe useful codecs Pareto Frontier is a chart where on one
3:19:41
Pareto Frontier is a chart where on one
3:19:41
Pareto Frontier is a chart where on one axis we have a compression ratio and
3:19:45
axis we have a compression ratio and
3:19:45
axis we have a compression ratio and another time
3:19:47
another time
3:19:47
another time typically time for CPU that is taken by
3:19:51
typically time for CPU that is taken by
3:19:51
typically time for CPU that is taken by compression
3:19:52
compression
3:19:52
compression and if some codec have no is some code
3:19:58
and if some codec have no is some code
3:19:58
and if some codec have no is some code have a data point which is not
3:19:59
have a data point which is not
3:19:59
have a data point which is not superseded by both
3:20:02
superseded by both
3:20:02
superseded by both access by some other codec by both
3:20:05
access by some other codec by both
3:20:05
access by some other codec by both access we say that the codec is on
3:20:07
access we say that the codec is on
3:20:07
access we say that the codec is on Pareto Frontier here we can see that on
3:20:11
Pareto Frontier here we can see that on
3:20:11
Pareto Frontier here we can see that on a higher uh timings and uh better
3:20:14
a higher uh timings and uh better
3:20:14
a higher uh timings and uh better compression ratios we usually have a
3:20:17
compression ratios we usually have a
3:20:17
compression ratios we usually have a lzma uh compression which is a lampard's
3:20:21
lzma uh compression which is a lampard's
3:20:21
lzma uh compression which is a lampard's event Markov chains algorithm and on the
3:20:25
event Markov chains algorithm and on the
3:20:25
event Markov chains algorithm and on the right side we have a faster codecs which
3:20:27
right side we have a faster codecs which
3:20:27
right side we have a faster codecs which consume a little CPU and have a modest
3:20:31
consume a little CPU and have a modest
3:20:31
consume a little CPU and have a modest compression ratio
3:20:33
compression ratio
3:20:33
compression ratio and so uh for our database we have some
3:20:38
and so uh for our database we have some
3:20:38
and so uh for our database we have some variable variable resources like memory
3:20:41
variable variable resources like memory
3:20:41
variable variable resources like memory but it's not affected much by
3:20:44
but it's not affected much by
3:20:44
but it's not affected much by compression and CPU time and the io time
3:20:49
compression and CPU time and the io time
3:20:49
compression and CPU time and the io time iOS throughput
3:20:52
iOS throughput
3:20:52
iOS throughput uh
3:20:54
uh
3:20:54
uh uh we have
3:20:55
uh we have
3:20:55
uh we have almost from the beginning of popularity
3:20:58
almost from the beginning of popularity
3:20:58
almost from the beginning of popularity of progress codec which is called glz uh
3:21:02
of progress codec which is called glz uh
3:21:02
of progress codec which is called glz uh this is a codec which was implemented by
3:21:04
this is a codec which was implemented by
3:21:04
this is a codec which was implemented by janowic back in
3:21:07
janowic back in
3:21:08
janowic back in 1997 and it was proposed in scientific
3:21:12
1997 and it was proposed in scientific
3:21:12
1997 and it was proposed in scientific paper as a as a simple compression
3:21:16
paper as a as a simple compression
3:21:16
paper as a as a simple compression algorithm back in 1993. so it's old it's
3:21:22
algorithm back in 1993. so it's old it's
3:21:22
algorithm back in 1993. so it's old it's not super efficient but the main uh uh
3:21:26
not super efficient but the main uh uh
3:21:26
not super efficient but the main uh uh like advantage of this codec was that
3:21:29
like advantage of this codec was that
3:21:29
like advantage of this codec was that it's it wasn't covered by any kind of
3:21:31
it's it wasn't covered by any kind of
3:21:31
it's it wasn't covered by any kind of patience so it could be used in
3:21:37
um in open source and this is this is a
3:21:40
um in open source and this is this is a
3:21:40
um in open source and this is this is a relatively important I think for for
3:21:45
relatively important I think for for
3:21:45
relatively important I think for for software
3:21:47
software
3:21:47
software uh later we have we had a new codec
3:21:51
uh later we have we had a new codec
3:21:51
uh later we have we had a new codec which is called zilip it was
3:21:54
which is called zilip it was
3:21:54
which is called zilip it was standardized by IDF with our RFC
3:21:58
standardized by IDF with our RFC
3:21:58
standardized by IDF with our RFC document uh it will it had a good
3:22:02
document uh it will it had a good
3:22:02
document uh it will it had a good implementation back in 2000s
3:22:05
implementation back in 2000s
3:22:05
implementation back in 2000s at the beginning of 2000 and from
3:22:10
at the beginning of 2000 and from
3:22:10
at the beginning of 2000 and from postgres is using zlib too it is
3:22:13
postgres is using zlib too it is
3:22:13
postgres is using zlib too it is slightly better both in a compression
3:22:16
slightly better both in a compression
3:22:16
slightly better both in a compression ratio and in compression performance
3:22:20
ratio and in compression performance
3:22:20
ratio and in compression performance uh but it was like not that importantly
3:22:24
uh but it was like not that importantly
3:22:24
uh but it was like not that importantly better than pglz to replace pglz
3:22:27
better than pglz to replace pglz
3:22:27
better than pglz to replace pglz everywhere
3:22:29
everywhere
3:22:29
everywhere uh later uh
3:22:33
uh later uh
3:22:33
uh later uh emerged another codec lz4 uh
3:22:38
emerged another codec lz4 uh
3:22:38
emerged another codec lz4 uh which is a compression optimized for
3:22:40
which is a compression optimized for
3:22:40
which is a compression optimized for performance it has a very modest
3:22:43
performance it has a very modest
3:22:43
performance it has a very modest compression ratio usually it compressed
3:22:45
compression ratio usually it compressed
3:22:45
compression ratio usually it compressed like
3:22:46
like
3:22:46
like normal date some average data to fold
3:22:50
normal date some average data to fold
3:22:50
normal date some average data to fold to compression ratio is around two but
3:22:55
to compression ratio is around two but
3:22:55
to compression ratio is around two but the main advantage of this codec is that
3:22:58
the main advantage of this codec is that
3:22:58
the main advantage of this codec is that the decompression
3:23:00
the decompression
3:23:00
the decompression almost with a speed of of memory so we
3:23:04
almost with a speed of of memory so we
3:23:04
almost with a speed of of memory so we see here as a just coping bytes is uh
3:23:08
see here as a just coping bytes is uh
3:23:09
see here as a just coping bytes is uh 14 gigabytes per CPU core and
3:23:11
14 gigabytes per CPU core and
3:23:11
14 gigabytes per CPU core and decompression lz4 is 5 gigabytes per CPU
3:23:16
decompression lz4 is 5 gigabytes per CPU
3:23:16
decompression lz4 is 5 gigabytes per CPU core which is a very impressive result
3:23:21
and later author of lz4 you can call it
3:23:25
and later author of lz4 you can call it
3:23:25
and later author of lz4 you can call it tried to supersede lz4 with uh Z
3:23:30
tried to supersede lz4 with uh Z
3:23:30
tried to supersede lz4 with uh Z standard codec uh which was expected to
3:23:34
standard codec uh which was expected to
3:23:34
standard codec uh which was expected to be
3:23:35
be
3:23:35
be better on Pareto Frontiers and than lz4
3:23:38
better on Pareto Frontiers and than lz4
3:23:38
better on Pareto Frontiers and than lz4 but it actually didn't lc4 is still
3:23:42
but it actually didn't lc4 is still
3:23:42
but it actually didn't lc4 is still super fast and the yesterday even with a
3:23:46
super fast and the yesterday even with a
3:23:46
super fast and the yesterday even with a lower compression uh level
3:23:49
lower compression uh level
3:23:49
lower compression uh level is not always faster than lz4
3:23:52
is not always faster than lz4
3:23:52
is not always faster than lz4 but this this standard is a main uh
3:23:59
but this this standard is a main uh
3:23:59
but this this standard is a main uh Frontier for database applications for
3:24:03
Frontier for database applications for
3:24:03
Frontier for database applications for database systems
3:24:05
database systems
3:24:05
database systems and is widely used
3:24:09
and is widely used
3:24:09
and is widely used and all these codecs are currently used
3:24:12
and all these codecs are currently used
3:24:12
and all these codecs are currently used in postgres let's see how they used ah
3:24:15
in postgres let's see how they used ah
3:24:15
in postgres let's see how they used ah now sorry now one more thing uh another
3:24:19
now sorry now one more thing uh another
3:24:19
now sorry now one more thing uh another thing that emerged relatively recently
3:24:21
thing that emerged relatively recently
3:24:21
thing that emerged relatively recently with codex that typically we have we
3:24:25
with codex that typically we have we
3:24:25
with codex that typically we have we need to provide code exam uh run away to
3:24:28
need to provide code exam uh run away to
3:24:28
need to provide code exam uh run away to unfold the com effective compression
3:24:31
unfold the com effective compression
3:24:31
unfold the com effective compression but in database uh
3:24:35
but in database uh
3:24:35
but in database uh from time to time we have to compress
3:24:37
from time to time we have to compress
3:24:37
from time to time we have to compress small buttons like strings of a few
3:24:41
small buttons like strings of a few
3:24:41
small buttons like strings of a few hundred bytes
3:24:43
hundred bytes
3:24:43
hundred bytes but for example uh z z standard and lz4
3:24:49
but for example uh z z standard and lz4
3:24:49
but for example uh z z standard and lz4 I almost indistinguishable in a
3:24:51
I almost indistinguishable in a
3:24:51
I almost indistinguishable in a compression ratio on sizes of a few
3:24:56
compression ratio on sizes of a few
3:24:56
compression ratio on sizes of a few kilobytes
3:24:57
kilobytes
3:24:57
kilobytes but to help with compression of small
3:25:00
but to help with compression of small
3:25:00
but to help with compression of small datums small data we can
3:25:04
datums small data we can
3:25:04
datums small data we can use compression dictionaries which is a
3:25:07
use compression dictionaries which is a
3:25:07
use compression dictionaries which is a prefix of like virtual prefix of every
3:25:11
prefix of like virtual prefix of every
3:25:11
prefix of like virtual prefix of every data that must be pre-computed for on
3:25:14
data that must be pre-computed for on
3:25:14
data that must be pre-computed for on some Corpus prepared before we compress
3:25:17
some Corpus prepared before we compress
3:25:17
some Corpus prepared before we compress anything and then it helps codec to
3:25:20
anything and then it helps codec to
3:25:21
anything and then it helps codec to compress uh frequent but small
3:25:24
compress uh frequent but small
3:25:24
compress uh frequent but small byte sequences
3:25:29
so all in all we see that codecs are
3:25:33
so all in all we see that codecs are
3:25:33
so all in all we see that codecs are became much better in recent times and
3:25:36
became much better in recent times and
3:25:36
became much better in recent times and here is a quote from Michelle pakia from
3:25:40
here is a quote from Michelle pakia from
3:25:40
here is a quote from Michelle pakia from pgsql hackers that modern compression
3:25:42
pgsql hackers that modern compression
3:25:42
pgsql hackers that modern compression algorithms became really useful for
3:25:45
algorithms became really useful for
3:25:45
algorithms became really useful for databases and I'd say that in last uh
3:25:49
databases and I'd say that in last uh
3:25:49
databases and I'd say that in last uh three or four years we made a huge
3:25:52
three or four years we made a huge
3:25:52
three or four years we made a huge advancement in postgres core to use
3:25:54
advancement in postgres core to use
3:25:54
advancement in postgres core to use better better compression algorithms
3:25:56
better better compression algorithms
3:25:56
better better compression algorithms let's see what's there and what still to
3:25:59
let's see what's there and what still to
3:25:59
let's see what's there and what still to do
3:26:00
do
3:26:00
do use cases in polgress first of all
3:26:03
use cases in polgress first of all
3:26:03
use cases in polgress first of all surely in Poltergeist 14 you can use
3:26:06
surely in Poltergeist 14 you can use
3:26:06
surely in Poltergeist 14 you can use better compression for toasts the other
3:26:09
better compression for toasts the other
3:26:09
better compression for toasts the other side
3:26:10
side
3:26:10
side attribute storage
3:26:12
attribute storage
3:26:12
attribute storage uh this is just a joke no don't use
3:26:16
uh this is just a joke no don't use
3:26:16
uh this is just a joke no don't use varcher to use text every everywhere
3:26:19
varcher to use text every everywhere
3:26:19
varcher to use text every everywhere text is just fine data type uh so
3:26:24
text is just fine data type uh so
3:26:24
text is just fine data type uh so most useful uh
3:26:29
most useful uh
3:26:29
most useful uh application of uh compression is a
3:26:32
application of uh compression is a
3:26:32
application of uh compression is a compression of toasts previously they
3:26:34
compression of toasts previously they
3:26:34
compression of toasts previously they were compressed with pglz now you can
3:26:38
were compressed with pglz now you can
3:26:38
were compressed with pglz now you can use it for uh text columns
3:26:43
use it for uh text columns
3:26:43
use it for uh text columns and when you create a
3:26:46
and when you create a
3:26:46
and when you create a column you can say that this call this
3:26:49
column you can say that this call this
3:26:49
column you can say that this call this column must be compressed with with lz4
3:26:52
column must be compressed with with lz4
3:26:52
column must be compressed with with lz4 unfortunately that's not a default so
3:26:55
unfortunately that's not a default so
3:26:55
unfortunately that's not a default so you have to alter table and set
3:26:59
you have to alter table and set
3:26:59
you have to alter table and set this compression yourself
3:27:04
default is still pglz
3:27:07
default is still pglz
3:27:07
default is still pglz but this setting will make a reading
3:27:11
but this setting will make a reading
3:27:11
but this setting will make a reading toasts much much faster on average lc4
3:27:16
toasts much much faster on average lc4
3:27:16
toasts much much faster on average lc4 is faster approximately paint like from
3:27:20
is faster approximately paint like from
3:27:20
is faster approximately paint like from five to ten times faster than
3:27:22
five to ten times faster than
3:27:22
five to ten times faster than decompression uh glz
3:27:26
another application of a compression is
3:27:29
another application of a compression is
3:27:29
another application of a compression is a base backup like PG based backup uh uh
3:27:33
a base backup like PG based backup uh uh
3:27:33
a base backup like PG based backup uh uh recently there was a new compression
3:27:37
recently there was a new compression
3:27:37
recently there was a new compression method added and you can say that
3:27:41
method added and you can say that
3:27:41
method added and you can say that compression must be done on a server
3:27:43
compression must be done on a server
3:27:43
compression must be done on a server what's the different what's the
3:27:45
what's the different what's the
3:27:45
what's the different what's the difference you can ask and the actual
3:27:48
difference you can ask and the actual
3:27:48
difference you can ask and the actual difference is that uh
3:27:51
difference is that uh
3:27:51
difference is that uh Network bandwidth is a super important
3:27:54
Network bandwidth is a super important
3:27:54
Network bandwidth is a super important resource
3:27:56
resource
3:27:56
resource when you are downloading a backup from
3:28:00
when you are downloading a backup from
3:28:00
when you are downloading a backup from running OTP installation you're
3:28:03
running OTP installation you're
3:28:03
running OTP installation you're consuming a network which is uh used for
3:28:08
consuming a network which is uh used for
3:28:08
consuming a network which is uh used for queries for replication for Waller
3:28:11
queries for replication for Waller
3:28:11
queries for replication for Waller hiding for lots of very important stuff
3:28:16
hiding for lots of very important stuff
3:28:16
hiding for lots of very important stuff and in many cases the limiting resource
3:28:21
and in many cases the limiting resource
3:28:21
and in many cases the limiting resource is not a
3:28:22
is not a
3:28:22
is not a [Music]
3:28:23
[Music]
3:28:23
[Music] um
3:28:24
um
3:28:24
um storage is not a storage quote or
3:28:26
storage is not a storage quote or
3:28:26
storage is not a storage quote or storage place on a disks starts empty
3:28:30
storage place on a disks starts empty
3:28:30
storage place on a disks starts empty space and disks but rather uh
3:28:34
space and disks but rather uh
3:28:34
space and disks but rather uh network throughput of uh highly loaded
3:28:37
network throughput of uh highly loaded
3:28:37
network throughput of uh highly loaded installation
3:28:39
installation
3:28:39
installation but actually there are so many other
3:28:42
but actually there are so many other
3:28:42
but actually there are so many other makeup tools like YG or PG Bucharest or
3:28:46
makeup tools like YG or PG Bucharest or
3:28:46
makeup tools like YG or PG Bucharest or Barman or
3:28:48
Barman or
3:28:48
Barman or uh I know
3:28:50
uh I know
3:28:50
uh I know G probe cup Armin and many many many
3:28:53
G probe cup Armin and many many many
3:28:53
G probe cup Armin and many many many others uh very good recap tools which
3:28:56
others uh very good recap tools which
3:28:56
others uh very good recap tools which can be used for backupin and they are
3:28:59
can be used for backupin and they are
3:28:59
can be used for backupin and they are actually already we're doing
3:29:02
actually already we're doing
3:29:02
actually already we're doing um compression
3:29:03
um compression
3:29:03
um compression of Base backup but anyway it's good to
3:29:07
of Base backup but anyway it's good to
3:29:07
of Base backup but anyway it's good to see a like good uh architectural
3:29:10
see a like good uh architectural
3:29:10
see a like good uh architectural decision it's in made in core and not in
3:29:14
decision it's in made in core and not in
3:29:14
decision it's in made in core and not in external tools
3:29:17
another interesting idea is the
3:29:20
another interesting idea is the
3:29:20
another interesting idea is the compression of wall right a head log uh
3:29:24
compression of wall right a head log uh
3:29:24
compression of wall right a head log uh let's see uh What uh resources wall can
3:29:28
let's see uh What uh resources wall can
3:29:28
let's see uh What uh resources wall can consume
3:29:29
consume
3:29:29
consume we see here that uh
3:29:32
we see here that uh
3:29:33
we see here that uh typical highly available postgres
3:29:35
typical highly available postgres
3:29:35
typical highly available postgres installation is sending wall to our Hive
3:29:39
installation is sending wall to our Hive
3:29:39
installation is sending wall to our Hive for a point in time recovery purposes it
3:29:44
for a point in time recovery purposes it
3:29:44
for a point in time recovery purposes it is reading wall from disks it's writing
3:29:47
is reading wall from disks it's writing
3:29:48
is reading wall from disks it's writing wall for crash recovery on primary it's
3:29:51
wall for crash recovery on primary it's
3:29:51
wall for crash recovery on primary it's sending wall to each of replica or to
3:29:54
sending wall to each of replica or to
3:29:54
sending wall to each of replica or to each uh standby instance and actually
3:29:58
each uh standby instance and actually
3:29:58
each uh standby instance and actually right in this wall on Eastern by
3:30:00
right in this wall on Eastern by
3:30:00
right in this wall on Eastern by instance on disk so if we
3:30:05
instance on disk so if we
3:30:05
instance on disk so if we uh for every single byte if we reduce uh
3:30:10
uh for every single byte if we reduce uh
3:30:10
uh for every single byte if we reduce uh [Music]
3:30:11
[Music]
3:30:11
[Music] amount of wool written a little uh we
3:30:15
amount of wool written a little uh we
3:30:15
amount of wool written a little uh we actually save a lot of resources in
3:30:17
actually save a lot of resources in
3:30:17
actually save a lot of resources in different places of a whole installation
3:30:20
different places of a whole installation
3:30:20
different places of a whole installation so it it kind of makes sense but
3:30:22
so it it kind of makes sense but
3:30:22
so it it kind of makes sense but compressing wool before writing it on
3:30:25
compressing wool before writing it on
3:30:25
compressing wool before writing it on this is a very tricky thing
3:30:28
this is a very tricky thing
3:30:28
this is a very tricky thing uh
3:30:32
um here are some problems of wall
3:30:35
um here are some problems of wall
3:30:35
um here are some problems of wall compression uh for example
3:30:39
compression uh for example
3:30:39
compression uh for example um actually all records are relatively
3:30:42
um actually all records are relatively
3:30:42
um actually all records are relatively small they're typically like on a scale
3:30:46
small they're typically like on a scale
3:30:46
small they're typically like on a scale of uh tens of bytes like smaller than
3:30:51
of uh tens of bytes like smaller than
3:30:51
of uh tens of bytes like smaller than 100 bytes and these chunks are not
3:30:54
100 bytes and these chunks are not
3:30:54
100 bytes and these chunks are not easily compressible even with more than
3:30:56
easily compressible even with more than
3:30:56
easily compressible even with more than modern algorithms but still they are
3:30:58
modern algorithms but still they are
3:30:58
modern algorithms but still they are compressible
3:31:00
compressible
3:31:00
compressible especially when they are going one after
3:31:03
especially when they are going one after
3:31:03
especially when they are going one after each other
3:31:04
each other
3:31:04
each other another problem is that if we compress
3:31:07
another problem is that if we compress
3:31:07
another problem is that if we compress wall before writing it on segments to
3:31:10
wall before writing it on segments to
3:31:10
wall before writing it on segments to disks to disk uh
3:31:14
disks to disk uh
3:31:14
disks to disk uh each wall sender will have to recompress
3:31:17
each wall sender will have to recompress
3:31:17
each wall sender will have to recompress it and while receiver have to compress
3:31:20
it and while receiver have to compress
3:31:20
it and while receiver have to compress it again
3:31:21
it again
3:31:21
it again and finally a crash recovery will be
3:31:24
and finally a crash recovery will be
3:31:24
and finally a crash recovery will be even more complicated than it is now and
3:31:27
even more complicated than it is now and
3:31:27
even more complicated than it is now and in fact a code of postgres crush
3:31:30
in fact a code of postgres crush
3:31:30
in fact a code of postgres crush recovery is uh
3:31:32
recovery is uh
3:31:33
recovery is uh very difficult for a new color to read
3:31:35
very difficult for a new color to read
3:31:35
very difficult for a new color to read it's super well documented so it's very
3:31:40
it's super well documented so it's very
3:31:40
it's super well documented so it's very well documented but
3:31:41
well documented but
3:31:41
well documented but it's simply complicated
3:31:44
it's simply complicated
3:31:44
it's simply complicated and in case of a compressed wall it will
3:31:49
and in case of a compressed wall it will
3:31:49
and in case of a compressed wall it will be much more complicated because we have
3:31:51
be much more complicated because we have
3:31:51
be much more complicated because we have to continue right in the same wall
3:31:53
to continue right in the same wall
3:31:53
to continue right in the same wall segment and modern algorithms
3:31:57
segment and modern algorithms
3:31:57
segment and modern algorithms simply do not allow you to add data to
3:32:01
simply do not allow you to add data to
3:32:01
simply do not allow you to add data to already compressed data you have to
3:32:03
already compressed data you have to
3:32:03
already compressed data you have to extract everything and compress it back
3:32:06
extract everything and compress it back
3:32:06
extract everything and compress it back with some suffix
3:32:09
with some suffix
3:32:09
with some suffix but
3:32:10
but
3:32:10
but if a very small compressed chunk can
3:32:14
if a very small compressed chunk can
3:32:14
if a very small compressed chunk can constitute a very large decompressed
3:32:16
constitute a very large decompressed
3:32:16
constitute a very large decompressed chunk and you even cannot predict how
3:32:19
chunk and you even cannot predict how
3:32:19
chunk and you even cannot predict how many memory you will need to decompress
3:32:21
many memory you will need to decompress
3:32:21
many memory you will need to decompress few bytes because every few bytes can be
3:32:25
few bytes because every few bytes can be
3:32:25
few bytes because every few bytes can be like a bombed over decompressed stuff
3:32:30
like a bombed over decompressed stuff
3:32:30
like a bombed over decompressed stuff anyway I think at some point we will
3:32:32
anyway I think at some point we will
3:32:32
anyway I think at some point we will have a wall compression because it's
3:32:35
have a wall compression because it's
3:32:35
have a wall compression because it's saves us a lot of
3:32:37
saves us a lot of
3:32:37
saves us a lot of computational resources a lot of these
3:32:40
computational resources a lot of these
3:32:40
computational resources a lot of these bandwidths and network bandwidth with a
3:32:43
bandwidths and network bandwidth with a
3:32:43
bandwidths and network bandwidth with a very tiny fraction of CPU time
3:32:47
very tiny fraction of CPU time
3:32:47
very tiny fraction of CPU time but already but but now we are already
3:32:50
but already but but now we are already
3:32:50
but already but but now we are already compressing full page images and
3:32:53
compressing full page images and
3:32:53
compressing full page images and starting from uh postgres 15 I as far as
3:32:58
starting from uh postgres 15 I as far as
3:32:58
starting from uh postgres 15 I as far as I remember uh
3:33:01
you already can use lz4 to compress full
3:33:05
you already can use lz4 to compress full
3:33:05
you already can use lz4 to compress full page images
3:33:06
page images
3:33:06
page images I contributed
3:33:08
I contributed
3:33:08
I contributed this patch together with Justin prisby
3:33:12
this patch together with Justin prisby
3:33:12
this patch together with Justin prisby and of course reviewers and help of
3:33:14
and of course reviewers and help of
3:33:14
and of course reviewers and help of community uh
3:33:17
according to my measurements it's me it
3:33:21
according to my measurements it's me it
3:33:21
according to my measurements it's me it makes average installation of uh like
3:33:25
makes average installation of uh like
3:33:25
makes average installation of uh like about 10 percent faster under PG bench
3:33:30
about 10 percent faster under PG bench
3:33:30
about 10 percent faster under PG bench and also you can use uh Z standard for
3:33:35
and also you can use uh Z standard for
3:33:35
and also you can use uh Z standard for for full page images compression too but
3:33:38
for full page images compression too but
3:33:38
for full page images compression too but on a scale of what single uh page image
3:33:42
on a scale of what single uh page image
3:33:42
on a scale of what single uh page image it I really did not observe real
3:33:46
it I really did not observe real
3:33:46
it I really did not observe real difference between these two codecs so
3:33:49
difference between these two codecs so
3:33:49
difference between these two codecs so they they both are much better than pglz
3:33:52
they they both are much better than pglz
3:33:52
they they both are much better than pglz but still uh
3:33:55
but still uh
3:33:55
but still uh we need to make something more clever
3:33:57
we need to make something more clever
3:33:57
we need to make something more clever with fpis to make uh this standard
3:34:02
with fpis to make uh this standard
3:34:02
with fpis to make uh this standard really show off I think that good idea
3:34:06
really show off I think that good idea
3:34:06
really show off I think that good idea would be to compress all page images
3:34:08
would be to compress all page images
3:34:08
would be to compress all page images together like having many page images
3:34:11
together like having many page images
3:34:11
together like having many page images that go one after each other they could
3:34:14
that go one after each other they could
3:34:14
that go one after each other they could be compressed together to make uh
3:34:17
be compressed together to make uh
3:34:17
be compressed together to make uh to make compression more efficient or we
3:34:19
to make compression more efficient or we
3:34:19
to make compression more efficient or we could try to keep a context of a
3:34:22
could try to keep a context of a
3:34:22
could try to keep a context of a compression between different
3:34:23
compression between different
3:34:23
compression between different compression Cycles
3:34:25
compression Cycles
3:34:25
compression Cycles but it will complicate a crush recovery
3:34:28
but it will complicate a crush recovery
3:34:28
but it will complicate a crush recovery so
3:34:30
so
3:34:30
so still straight off of complexity and
3:34:33
still straight off of complexity and
3:34:33
still straight off of complexity and efficiency not just of uh
3:34:35
efficiency not just of uh
3:34:36
efficiency not just of uh [Music]
3:34:37
[Music]
3:34:37
[Music] compression ratio and compression time
3:34:41
compression ratio and compression time
3:34:41
compression ratio and compression time also communities working on protocol
3:34:43
also communities working on protocol
3:34:43
also communities working on protocol compression like compression of flip EQ
3:34:45
compression like compression of flip EQ
3:34:45
compression like compression of flip EQ protocol
3:34:46
protocol
3:34:46
protocol uh it is sometimes the thread is
3:34:50
uh it is sometimes the thread is
3:34:50
uh it is sometimes the thread is resurrected sometimes it's returned with
3:34:52
resurrected sometimes it's returned with
3:34:52
resurrected sometimes it's returned with feedback and currently the main concern
3:34:55
feedback and currently the main concern
3:34:55
feedback and currently the main concern is that compression is in fact defeating
3:34:59
is that compression is in fact defeating
3:34:59
is that compression is in fact defeating TLS we fit in encryption
3:35:03
TLS we fit in encryption
3:35:03
TLS we fit in encryption um
3:35:04
um
3:35:04
um what's the problem here let's see what
3:35:07
what's the problem here let's see what
3:35:07
what's the problem here let's see what is uh crime attack on a compression uh
3:35:11
is uh crime attack on a compression uh
3:35:11
is uh crime attack on a compression uh before encryption if we compress
3:35:14
before encryption if we compress
3:35:14
before encryption if we compress together some secret data with some user
3:35:17
together some secret data with some user
3:35:17
together some secret data with some user control data
3:35:18
control data
3:35:18
control data the whole idea of
3:35:21
the whole idea of
3:35:21
the whole idea of compression is making some more frequent
3:35:26
compression is making some more frequent
3:35:26
compression is making some more frequent data sequences shorter but how do we
3:35:29
data sequences shorter but how do we
3:35:29
data sequences shorter but how do we know that some uh some
3:35:32
know that some uh some
3:35:32
know that some uh some [Music]
3:35:34
[Music]
3:35:34
[Music] consequence is more frequent
3:35:36
consequence is more frequent
3:35:36
consequence is more frequent the answer is simple if uh
3:35:41
the answer is simple if uh
3:35:41
the answer is simple if uh if the byte sequence is self-resemblant
3:35:44
if the byte sequence is self-resemblant
3:35:45
if the byte sequence is self-resemblant if it's if it is repeating itself then
3:35:47
if it's if it is repeating itself then
3:35:48
if it's if it is repeating itself then we think this is a like more frequent so
3:35:51
we think this is a like more frequent so
3:35:51
we think this is a like more frequent so any compression is uh
3:35:55
any compression is uh
3:35:55
any compression is uh is working better if the data resembles
3:35:59
is working better if the data resembles
3:35:59
is working better if the data resembles itself and if we have a compressor you
3:36:03
itself and if we have a compressor you
3:36:03
itself and if we have a compressor you secret data with user control data user
3:36:07
secret data with user control data user
3:36:07
secret data with user control data user can judge uh can extract information on
3:36:11
can judge uh can extract information on
3:36:11
can judge uh can extract information on the resemblance between secret data and
3:36:14
the resemblance between secret data and
3:36:14
the resemblance between secret data and user data and if the user is observing a
3:36:18
user data and if the user is observing a
3:36:18
user data and if the user is observing a size of encrypted data on public Channel
3:36:22
size of encrypted data on public Channel
3:36:22
size of encrypted data on public Channel they can judge how efficient it how uh
3:36:26
they can judge how efficient it how uh
3:36:26
they can judge how efficient it how uh what what is actually the secret data
3:36:29
what what is actually the secret data
3:36:29
what what is actually the secret data and if attacker can repeat again and
3:36:31
and if attacker can repeat again and
3:36:31
and if attacker can repeat again and again the same query
3:36:34
again the same query
3:36:34
again the same query they can eventually extract secret data
3:36:37
they can eventually extract secret data
3:36:37
they can eventually extract secret data from encrypted Channel
3:36:39
from encrypted Channel
3:36:39
from encrypted Channel it's very theoretical attack but
3:36:42
it's very theoretical attack but
3:36:42
it's very theoretical attack but plausible that's why we don't have a
3:36:45
plausible that's why we don't have a
3:36:45
plausible that's why we don't have a compression in open necessary
3:36:47
compression in open necessary
3:36:47
compression in open necessary implementation and other TLS
3:36:49
implementation and other TLS
3:36:49
implementation and other TLS implementations
3:36:50
implementations
3:36:50
implementations but still compression of protocol will
3:36:54
but still compression of protocol will
3:36:54
but still compression of protocol will save us a lot of resources so at some
3:36:56
save us a lot of resources so at some
3:36:56
save us a lot of resources so at some point we will have to have it
3:36:58
point we will have to have it
3:36:58
point we will have to have it despite of some security risks
3:37:02
another wild idea of fusion compression
3:37:05
another wild idea of fusion compression
3:37:05
another wild idea of fusion compression is compression of temporary files uh
3:37:08
is compression of temporary files uh
3:37:08
is compression of temporary files uh like temporary files will go through iOS
3:37:11
like temporary files will go through iOS
3:37:11
like temporary files will go through iOS that's why they deserve a compression I
3:37:14
that's why they deserve a compression I
3:37:14
that's why they deserve a compression I have hacked the patch here is a like
3:37:17
have hacked the patch here is a like
3:37:17
have hacked the patch here is a like just a
3:37:18
just a
3:37:18
just a screenshot of postgresh Hacker's message
3:37:21
screenshot of postgresh Hacker's message
3:37:21
screenshot of postgresh Hacker's message about this patch
3:37:24
it somehow works and allows uh to reduce
3:37:28
it somehow works and allows uh to reduce
3:37:28
it somehow works and allows uh to reduce uh four times number of bytes written
3:37:32
uh four times number of bytes written
3:37:32
uh four times number of bytes written during create index separation or hash
3:37:35
during create index separation or hash
3:37:35
during create index separation or hash join operation
3:37:37
join operation
3:37:37
join operation but
3:37:39
but
3:37:39
but the implementations is just just too
3:37:41
the implementations is just just too
3:37:41
the implementations is just just too difficult to just to complicate it and
3:37:44
difficult to just to complicate it and
3:37:44
difficult to just to complicate it and not usually worth it like
3:37:48
not usually worth it like
3:37:48
not usually worth it like um
3:37:48
um
3:37:48
um it's always com like trade off between
3:37:51
it's always com like trade off between
3:37:51
it's always com like trade off between maintainability and uh
3:37:55
profit for end users and so far the
3:37:59
profit for end users and so far the
3:37:59
profit for end users and so far the patch is not in the shape what's the
3:38:02
patch is not in the shape what's the
3:38:02
patch is not in the shape what's the basic problem of this uh patch uh it's
3:38:05
basic problem of this uh patch uh it's
3:38:05
basic problem of this uh patch uh it's based on the idea of Random Access
3:38:07
based on the idea of Random Access
3:38:07
based on the idea of Random Access compressed file and each page after
3:38:10
compressed file and each page after
3:38:10
compressed file and each page after compression have a different size it can
3:38:13
compression have a different size it can
3:38:13
compression have a different size it can be
3:38:15
be
3:38:15
be uh wide resonant was more than eight
3:38:17
uh wide resonant was more than eight
3:38:17
uh wide resonant was more than eight kilobytes it can be slower it can be
3:38:20
kilobytes it can be slower it can be
3:38:20
kilobytes it can be slower it can be smaller than it was and at some point
3:38:23
smaller than it was and at some point
3:38:23
smaller than it was and at some point pages will be deleted and
3:38:26
pages will be deleted and
3:38:26
pages will be deleted and actually defragmented temporary file is
3:38:29
actually defragmented temporary file is
3:38:30
actually defragmented temporary file is like sounds like nonsense but still when
3:38:34
like sounds like nonsense but still when
3:38:34
like sounds like nonsense but still when we write something in the middle of a
3:38:37
we write something in the middle of a
3:38:37
we write something in the middle of a temporary file we will have to
3:38:40
temporary file we will have to
3:38:40
temporary file we will have to defragment it somehow and this complex
3:38:43
defragment it somehow and this complex
3:38:43
defragment it somehow and this complex tip currently prevents me from uh going
3:38:47
tip currently prevents me from uh going
3:38:47
tip currently prevents me from uh going further with this technology
3:38:50
further with this technology
3:38:50
further with this technology there are some Alternatives available
3:38:54
there are some Alternatives available
3:38:54
there are some Alternatives available for example in postgres Pro uh it's it's
3:38:58
for example in postgres Pro uh it's it's
3:38:58
for example in postgres Pro uh it's it's a proprietary Fork closed Source but
3:39:01
a proprietary Fork closed Source but
3:39:01
a proprietary Fork closed Source but they have like compressed file system
3:39:03
they have like compressed file system
3:39:03
they have like compressed file system compressed
3:39:06
um or maybe it's not complete I know
3:39:09
um or maybe it's not complete I know
3:39:09
um or maybe it's not complete I know they they call this technology CFS and
3:39:12
they they call this technology CFS and
3:39:12
they they call this technology CFS and this is effectively compression uh per
3:39:15
this is effectively compression uh per
3:39:15
this is effectively compression uh per page it I know that it works sometimes
3:39:18
page it I know that it works sometimes
3:39:18
page it I know that it works sometimes on some workloads uh very well on some
3:39:22
on some workloads uh very well on some
3:39:22
on some workloads uh very well on some workloads another option is just to use
3:39:25
workloads another option is just to use
3:39:25
workloads another option is just to use a compressing file system like the the
3:39:28
a compressing file system like the the
3:39:28
a compressing file system like the the first btrfs or some others
3:39:32
first btrfs or some others
3:39:32
first btrfs or some others uh another like develop develop
3:39:37
uh another like develop develop
3:39:37
uh another like develop develop development direction is to just go on
3:39:39
development direction is to just go on
3:39:39
development direction is to just go on with a defragmentation of Random Access
3:39:42
with a defragmentation of Random Access
3:39:42
with a defragmentation of Random Access compressed file but the problem is that
3:39:45
compressed file but the problem is that
3:39:45
compressed file but the problem is that uh we actually have to do it durable and
3:39:48
uh we actually have to do it durable and
3:39:48
uh we actually have to do it durable and if we want a durable defragmentation of
3:39:52
if we want a durable defragmentation of
3:39:52
if we want a durable defragmentation of temporary file we will have to log
3:39:55
temporary file we will have to log
3:39:55
temporary file we will have to log temporary file and this is just I'm
3:39:58
temporary file and this is just I'm
3:39:58
temporary file and this is just I'm hesitant to go go go on with this and
3:40:03
hesitant to go go go on with this and
3:40:03
hesitant to go go go on with this and another interesting idea is a approach
3:40:05
another interesting idea is a approach
3:40:06
another interesting idea is a approach taken by a green plum team they
3:40:08
taken by a green plum team they
3:40:08
taken by a green plum team they identified they had that hash join do
3:40:11
identified they had that hash join do
3:40:11
identified they had that hash join do not require Random Access it's always
3:40:14
not require Random Access it's always
3:40:14
not require Random Access it's always sequential and they extended buffer file
3:40:17
sequential and they extended buffer file
3:40:17
sequential and they extended buffer file API with a pledge sequential method that
3:40:20
API with a pledge sequential method that
3:40:20
API with a pledge sequential method that is saying that system profile will be
3:40:25
is saying that system profile will be
3:40:25
is saying that system profile will be never read from in in the middle never
3:40:27
never read from in in the middle never
3:40:27
never read from in in the middle never written in the middle just compress it
3:40:30
written in the middle just compress it
3:40:30
written in the middle just compress it as sequential
3:40:31
as sequential
3:40:31
as sequential and it is uh usable but I don't think
3:40:35
and it is uh usable but I don't think
3:40:35
and it is uh usable but I don't think that patch as it is in uh green plum is
3:40:38
that patch as it is in uh green plum is
3:40:38
that patch as it is in uh green plum is in any close to commutable uh in
3:40:41
in any close to commutable uh in
3:40:41
in any close to commutable uh in postgres right now
3:40:43
postgres right now
3:40:43
postgres right now so if you want to remember something
3:40:46
so if you want to remember something
3:40:46
so if you want to remember something from this talk uh
3:40:49
from this talk uh
3:40:49
from this talk uh first thing that we're trying to
3:40:52
first thing that we're trying to
3:40:52
first thing that we're trying to compress frequent data and uh
3:40:56
compress frequent data and uh
3:40:56
compress frequent data and uh rare data can be longer than uh com than
3:41:00
rare data can be longer than uh com than
3:41:00
rare data can be longer than uh com than before compression and the other ideas
3:41:02
before compression and the other ideas
3:41:02
before compression and the other ideas that any bytes that go slow through goes
3:41:05
that any bytes that go slow through goes
3:41:05
that any bytes that go slow through goes to this code the network deserves a
3:41:07
to this code the network deserves a
3:41:07
to this code the network deserves a compression if you have some good ideas
3:41:10
compression if you have some good ideas
3:41:10
compression if you have some good ideas on
3:41:12
on
3:41:12
on compression in database I'd be happy to
3:41:14
compression in database I'd be happy to
3:41:14
compression in database I'd be happy to discuss it with you in Discord
3:41:18
discuss it with you in Discord
3:41:18
discuss it with you in Discord um
3:41:18
um
3:41:18
um also maybe you have some scientific
3:41:21
also maybe you have some scientific
3:41:21
also maybe you have some scientific ideas how to make a recompression in
3:41:24
ideas how to make a recompression in
3:41:24
ideas how to make a recompression in popular algorithms and if you are
3:41:27
popular algorithms and if you are
3:41:27
popular algorithms and if you are willing to work on lz4 and this they
3:41:30
willing to work on lz4 and this they
3:41:30
willing to work on lz4 and this they don't have this API that would be great
3:41:32
don't have this API that would be great
3:41:32
don't have this API that would be great uh
3:41:33
uh
3:41:33
uh [Music]
3:41:36
[Music]
3:41:36
[Music] and basically that's it I'll be happy to
3:41:38
and basically that's it I'll be happy to
3:41:38
and basically that's it I'll be happy to discuss is that this stuff with you in
3:41:41
discuss is that this stuff with you in
3:41:41
discuss is that this stuff with you in Discord thank you so much for listening
3:41:45
Discord thank you so much for listening
3:41:45
Discord thank you so much for listening yeah all right
3:41:47
yeah all right
3:41:48
yeah all right thank you Andre
3:41:50
thank you Andre
3:41:50
thank you Andre that was pretty good A lot of
3:41:52
that was pretty good A lot of
3:41:52
that was pretty good A lot of information packed into a short period
3:41:55
information packed into a short period
3:41:55
information packed into a short period of time so nicely done is that a pun is
3:41:59
of time so nicely done is that a pun is
3:41:59
of time so nicely done is that a pun is that a pun Rob come on maybe it could be
3:42:02
that a pun Rob come on maybe it could be
3:42:02
that a pun Rob come on maybe it could be late in the day could be a pun uh I did
3:42:05
late in the day could be a pun uh I did
3:42:05
late in the day could be a pun uh I did want to ask a couple of cool questions
3:42:06
want to ask a couple of cool questions
3:42:06
want to ask a couple of cool questions if you don't mind
3:42:08
if you don't mind
3:42:08
if you don't mind um so one for sure is uh I I think most
3:42:12
um so one for sure is uh I I think most
3:42:12
um so one for sure is uh I I think most of the research essentially you touched
3:42:14
of the research essentially you touched
3:42:14
of the research essentially you touched on some research that is not from
3:42:15
on some research that is not from
3:42:15
on some research that is not from Academia I always feel like most of it
3:42:17
Academia I always feel like most of it
3:42:17
Academia I always feel like most of it comes from Academia and some of it is
3:42:19
comes from Academia and some of it is
3:42:19
comes from Academia and some of it is even papers that were written like you
3:42:20
even papers that were written like you
3:42:20
even papers that were written like you know 20 years ago that we just never
3:42:22
know 20 years ago that we just never
3:42:22
know 20 years ago that we just never looked at for some reason
3:42:24
looked at for some reason
3:42:24
looked at for some reason um I'm curious like how do you feel
3:42:25
um I'm curious like how do you feel
3:42:25
um I'm curious like how do you feel about the state of taking that
3:42:27
about the state of taking that
3:42:27
about the state of taking that information and then being able to
3:42:29
information and then being able to
3:42:29
information and then being able to transform it into a way that's useful
3:42:31
transform it into a way that's useful
3:42:31
transform it into a way that's useful for production systems both like there's
3:42:33
for production systems both like there's
3:42:33
for production systems both like there's a technical issue there I'm sure but
3:42:35
a technical issue there I'm sure but
3:42:35
a technical issue there I'm sure but also like legal ones around like patents
3:42:37
also like legal ones around like patents
3:42:37
also like legal ones around like patents and that kind of stuff like any any
3:42:38
and that kind of stuff like any any
3:42:38
and that kind of stuff like any any thoughts on that
3:42:41
thoughts on that
3:42:41
thoughts on that mm-hmm yeah well a lot of very good
3:42:46
mm-hmm yeah well a lot of very good
3:42:46
mm-hmm yeah well a lot of very good research was done in uh in the times
3:42:49
research was done in uh in the times
3:42:49
research was done in uh in the times when parents were a thing and we cannot
3:42:52
when parents were a thing and we cannot
3:42:52
when parents were a thing and we cannot just use anything that is potent because
3:42:55
just use anything that is potent because
3:42:55
just use anything that is potent because like
3:42:57
like
3:42:57
like slightly it's a project policy and
3:43:01
slightly it's a project policy and
3:43:01
slightly it's a project policy and another thing is Academia is like doing
3:43:04
another thing is Academia is like doing
3:43:04
another thing is Academia is like doing a job well in a like bro on a broad
3:43:07
a job well in a like bro on a broad
3:43:07
a job well in a like bro on a broad scale they see us like database
3:43:10
scale they see us like database
3:43:10
scale they see us like database developers as just one application and
3:43:12
developers as just one application and
3:43:12
developers as just one application and they are producing algorithms that are
3:43:15
they are producing algorithms that are
3:43:15
they are producing algorithms that are like general purpose algorithms uh yeah
3:43:18
like general purpose algorithms uh yeah
3:43:18
like general purpose algorithms uh yeah they feel that we could have a
3:43:20
they feel that we could have a
3:43:20
they feel that we could have a algorithms that better see it uh our
3:43:23
algorithms that better see it uh our
3:43:23
algorithms that better see it uh our needs in databases and if some
3:43:26
needs in databases and if some
3:43:26
needs in databases and if some researchers are watching us so I'm
3:43:28
researchers are watching us so I'm
3:43:28
researchers are watching us so I'm asking you to think more about how how
3:43:31
asking you to think more about how how
3:43:31
asking you to think more about how how uh
3:43:32
uh
3:43:33
uh how we can adapt API because uh in
3:43:36
how we can adapt API because uh in
3:43:36
how we can adapt API because uh in databases we usually have a Pages we
3:43:39
databases we usually have a Pages we
3:43:39
databases we usually have a Pages we usually have uh like it we usually have
3:43:43
usually have uh like it we usually have
3:43:43
usually have uh like it we usually have something like tuples we usually have to
3:43:46
something like tuples we usually have to
3:43:46
something like tuples we usually have to add some data into already compression
3:43:48
add some data into already compression
3:43:48
add some data into already compression data and we have so many like ways to
3:43:53
data and we have so many like ways to
3:43:53
data and we have so many like ways to test uh compression algorithms which see
3:43:56
test uh compression algorithms which see
3:43:56
test uh compression algorithms which see it our needs better and we have a
3:43:58
it our needs better and we have a
3:43:58
it our needs better and we have a standard ways to measure performance of
3:44:01
standard ways to measure performance of
3:44:01
standard ways to measure performance of overall system not just a compression
3:44:02
overall system not just a compression
3:44:03
overall system not just a compression system if if your algorithm is uh like
3:44:07
system if if your algorithm is uh like
3:44:07
system if if your algorithm is uh like pressing uh
3:44:09
pressing uh
3:44:09
pressing uh CPU in some extraordinary way we will
3:44:12
CPU in some extraordinary way we will
3:44:12
CPU in some extraordinary way we will see it in a system working as a whole uh
3:44:16
see it in a system working as a whole uh
3:44:16
see it in a system working as a whole uh not not just a one isolated part so yeah
3:44:20
not not just a one isolated part so yeah
3:44:20
not not just a one isolated part so yeah it would be great if some researchers
3:44:22
it would be great if some researchers
3:44:22
it would be great if some researchers would join this uh with together with
3:44:25
would join this uh with together with
3:44:25
would join this uh with together with database scientists and the database
3:44:27
database scientists and the database
3:44:27
database scientists and the database developers the code is there they should
3:44:29
developers the code is there they should
3:44:29
developers the code is there they should join us I agree totally a call to action
3:44:33
join us I agree totally a call to action
3:44:33
join us I agree totally a call to action absolutely go to research
3:44:36
absolutely go to research
3:44:36
absolutely go to research so um we'd love to talk to you more I'm
3:44:40
so um we'd love to talk to you more I'm
3:44:40
so um we'd love to talk to you more I'm sure people have more questions there's
3:44:41
sure people have more questions there's
3:44:41
sure people have more questions there's probably some on the Discord if you
3:44:43
probably some on the Discord if you
3:44:43
probably some on the Discord if you you're going to be popping over there
3:44:46
you're going to be popping over there
3:44:46
you're going to be popping over there yeah sure all right and are you going to
3:44:48
yeah sure all right and are you going to
3:44:48
yeah sure all right and are you going to be at the emea live stream which is
3:44:50
be at the emea live stream which is
3:44:50
be at the emea live stream which is happening
3:44:51
happening
3:44:51
happening um 9 A.M central European summer time uh
3:44:56
um 9 A.M central European summer time uh
3:44:56
um 9 A.M central European summer time uh tomorrow yeah I will try I I'm planning
3:44:59
tomorrow yeah I will try I I'm planning
3:44:59
tomorrow yeah I will try I I'm planning to watch uh you know Marcus lot keynote
3:45:03
to watch uh you know Marcus lot keynote
3:45:03
to watch uh you know Marcus lot keynote and probably I will stay for some other
3:45:06
and probably I will stay for some other
3:45:06
and probably I will stay for some other talks too
3:45:07
talks too
3:45:07
talks too awesome well thank you again Andre for
3:45:10
awesome well thank you again Andre for
3:45:10
awesome well thank you again Andre for being part of this and um your talk it
3:45:13
being part of this and um your talk it
3:45:13
being part of this and um your talk it will be available for people who want to
3:45:15
will be available for people who want to
3:45:15
will be available for people who want to watch the replay of the America's live
3:45:17
watch the replay of the America's live
3:45:17
watch the replay of the America's live stream maybe if they got here late
3:45:19
stream maybe if they got here late
3:45:19
stream maybe if they got here late um they'll be able to watch it there and
3:45:20
um they'll be able to watch it there and
3:45:20
um they'll be able to watch it there and then we'll be publishing it as its own
3:45:22
then we'll be publishing it as its own
3:45:22
then we'll be publishing it as its own independent talk um on YouTube within
3:45:25
independent talk um on YouTube within
3:45:25
independent talk um on YouTube within the next couple of weeks cool thank you
3:45:27
the next couple of weeks cool thank you
3:45:27
the next couple of weeks cool thank you all right thanks Andre thank you so now
3:45:30
all right thanks Andre thank you so now
3:45:30
all right thanks Andre thank you so now it's time to go to the wrap-up and it's
3:45:35
it's time to go to the wrap-up and it's
3:45:35
it's time to go to the wrap-up and it's it's you and me
3:45:37
it's you and me
3:45:37
it's you and me um our wrap up is supposed to be like 25
3:45:38
um our wrap up is supposed to be like 25
3:45:38
um our wrap up is supposed to be like 25 minutes but we're gonna have to go
3:45:40
minutes but we're gonna have to go
3:45:40
minutes but we're gonna have to go faster up yeah you know I was gonna say
3:45:43
faster up yeah you know I was gonna say
3:45:43
faster up yeah you know I was gonna say like I can't tell if it was a long day
3:45:44
like I can't tell if it was a long day
3:45:44
like I can't tell if it was a long day or like a short day because it seemed
3:45:46
or like a short day because it seemed
3:45:46
or like a short day because it seemed like it went by pretty quick so but
3:45:48
like it went by pretty quick so but
3:45:48
like it went by pretty quick so but that's good that's good it did go by
3:45:50
that's good that's good it did go by
3:45:50
that's good that's good it did go by really quick and um especially paying
3:45:53
really quick and um especially paying
3:45:53
really quick and um especially paying attention to the chat and trying to pay
3:45:55
attention to the chat and trying to pay
3:45:55
attention to the chat and trying to pay attention to the talks and I forgot
3:45:57
attention to the talks and I forgot
3:45:57
attention to the talks and I forgot about live tweeting thank goodness that
3:45:59
about live tweeting thank goodness that
3:45:59
about live tweeting thank goodness that other people are out there out there
3:46:01
other people are out there out there
3:46:01
other people are out there out there doing that so we have a couple things um
3:46:03
doing that so we have a couple things um
3:46:03
doing that so we have a couple things um that we wanted to cover
3:46:05
that we wanted to cover
3:46:05
that we wanted to cover um in the beginning so let's see the
3:46:07
um in the beginning so let's see the
3:46:07
um in the beginning so let's see the first is just that reminder that there's
3:46:09
first is just that reminder that there's
3:46:09
first is just that reminder that there's a survey and uh you can quickly scan
3:46:12
a survey and uh you can quickly scan
3:46:12
a survey and uh you can quickly scan that QR code you can enter the survey
3:46:15
that QR code you can enter the survey
3:46:15
that QR code you can enter the survey feedback multiple times so if you
3:46:17
feedback multiple times so if you
3:46:17
feedback multiple times so if you haven't watched all the talks yet you
3:46:18
haven't watched all the talks yet you
3:46:18
haven't watched all the talks yet you want to watch some of the on-demand ones
3:46:20
want to watch some of the on-demand ones
3:46:20
want to watch some of the on-demand ones or whatever you can do it multiple times
3:46:22
or whatever you can do it multiple times
3:46:22
or whatever you can do it multiple times until it closes next Friday so yeah
3:46:24
until it closes next Friday so yeah
3:46:24
until it closes next Friday so yeah often vote early vote often that works
3:46:29
often vote early vote often that works
3:46:29
often vote early vote often that works um I I feel like it's a broken record
3:46:31
um I I feel like it's a broken record
3:46:31
um I I feel like it's a broken record I've mentioned the Discord um teen times
3:46:32
I've mentioned the Discord um teen times
3:46:32
I've mentioned the Discord um teen times so
3:46:34
so
3:46:34
so um there's a word for when you say a
3:46:36
um there's a word for when you say a
3:46:36
um there's a word for when you say a word so many times that it loses its its
3:46:38
word so many times that it loses its its
3:46:38
word so many times that it loses its its meaning inside your head
3:46:41
meaning inside your head
3:46:41
meaning inside your head no and actually that word was on a Ted
3:46:44
no and actually that word was on a Ted
3:46:44
no and actually that word was on a Ted lasso episode recently but that's that's
3:46:45
lasso episode recently but that's that's
3:46:45
lasso episode recently but that's that's another story and I just can't recall
3:46:47
another story and I just can't recall
3:46:47
another story and I just can't recall what it was but anyway I have said the
3:46:49
what it was but anyway I have said the
3:46:49
what it was but anyway I have said the word Discord so many times today but
3:46:52
word Discord so many times today but
3:46:52
word Discord so many times today but there is there is a kind of a fun back
3:46:54
there is there is a kind of a fun back
3:46:54
there is there is a kind of a fun back Channel there for those who want to join
3:46:56
Channel there for those who want to join
3:46:56
Channel there for those who want to join and it's going to continue throughout
3:46:57
and it's going to continue throughout
3:46:57
and it's going to continue throughout the rest of the day
3:46:59
the rest of the day
3:46:59
the rest of the day um
3:47:00
um
3:47:00
um wow big thank you to all of these
3:47:03
wow big thank you to all of these
3:47:03
wow big thank you to all of these speakers this is pretty cool and I have
3:47:06
speakers this is pretty cool and I have
3:47:06
speakers this is pretty cool and I have the URL showing for the live stream
3:47:08
the URL showing for the live stream
3:47:08
the URL showing for the live stream Playlist cytuscon Live stream playlist
3:47:10
Playlist cytuscon Live stream playlist
3:47:10
Playlist cytuscon Live stream playlist right now and that's why I pop this up
3:47:12
right now and that's why I pop this up
3:47:12
right now and that's why I pop this up there
3:47:13
there
3:47:13
there um
3:47:14
um
3:47:14
um and then if you want to mark your
3:47:16
and then if you want to mark your
3:47:16
and then if you want to mark your calendar I want to make it easy for you
3:47:18
calendar I want to make it easy for you
3:47:18
calendar I want to make it easy for you if you want to go to all or part of the
3:47:20
if you want to go to all or part of the
3:47:20
if you want to go to all or part of the emea live stream uh because who knows
3:47:22
emea live stream uh because who knows
3:47:22
emea live stream uh because who knows where you are in the world you might be
3:47:24
where you are in the world you might be
3:47:24
where you are in the world you might be watching this in Replay in a couple
3:47:26
watching this in Replay in a couple
3:47:26
watching this in Replay in a couple hours and maybe you live in Europe
3:47:28
hours and maybe you live in Europe
3:47:28
hours and maybe you live in Europe um so the link in the upper right corner
3:47:31
um so the link in the upper right corner
3:47:31
um so the link in the upper right corner aka.mscon hyphen India that gives you an
3:47:35
aka.mscon hyphen India that gives you an
3:47:35
aka.mscon hyphen India that gives you an easy way to drop something in your
3:47:36
easy way to drop something in your
3:47:36
easy way to drop something in your calendar for the Amia live stream
3:47:38
calendar for the Amia live stream
3:47:38
calendar for the Amia live stream tomorrow
3:47:39
tomorrow
3:47:39
tomorrow you'll be a sleep Rob I hope so I mean I
3:47:42
you'll be a sleep Rob I hope so I mean I
3:47:42
you'll be a sleep Rob I hope so I mean I do want to watch this I will definitely
3:47:44
do want to watch this I will definitely
3:47:44
do want to watch this I will definitely watch the replay of it but I don't think
3:47:46
watch the replay of it but I don't think
3:47:46
watch the replay of it but I don't think I will watch it live I think it will
3:47:48
I will watch it live I think it will
3:47:48
I will watch it live I think it will have gone really wrong if that happens
3:47:49
have gone really wrong if that happens
3:47:49
have gone really wrong if that happens so
3:47:52
so
3:47:52
so on-demand talks there's 25 of them
3:47:55
on-demand talks there's 25 of them
3:47:55
on-demand talks there's 25 of them that's that's a fair number of talks but
3:47:58
that's that's a fair number of talks but
3:47:58
that's that's a fair number of talks but they're all really well I don't know I
3:48:00
they're all really well I don't know I
3:48:00
they're all really well I don't know I assume they're all really good because
3:48:01
assume they're all really good because
3:48:01
assume they're all really good because I've looked through the list I've been
3:48:03
I've looked through the list I've been
3:48:03
I've looked through the list I've been trying to also produce that along with
3:48:05
trying to also produce that along with
3:48:05
trying to also produce that along with these other things we have going on
3:48:06
these other things we have going on
3:48:06
these other things we have going on today during the live stream was like
3:48:07
today during the live stream was like
3:48:07
today during the live stream was like looking through that list of talks and
3:48:09
looking through that list of talks and
3:48:09
looking through that list of talks and seeing what was there and uh I'm really
3:48:12
seeing what was there and uh I'm really
3:48:12
seeing what was there and uh I'm really excited actually I will be watching some
3:48:13
excited actually I will be watching some
3:48:13
excited actually I will be watching some of those tonight before you know before
3:48:15
of those tonight before you know before
3:48:15
of those tonight before you know before things happen tomorrow so for my tonight
3:48:18
things happen tomorrow so for my tonight
3:48:18
things happen tomorrow so for my tonight well and you gave a really interesting
3:48:21
well and you gave a really interesting
3:48:21
well and you gave a really interesting on-demand talk last year and I know that
3:48:24
on-demand talk last year and I know that
3:48:24
on-demand talk last year and I know that you know those views they don't all
3:48:25
you know those views they don't all
3:48:25
you know those views they don't all happen on the day of the event right
3:48:27
happen on the day of the event right
3:48:27
happen on the day of the event right they happen over time there's a whole
3:48:28
they happen over time there's a whole
3:48:28
they happen over time there's a whole long tail and so the good news is people
3:48:30
long tail and so the good news is people
3:48:30
long tail and so the good news is people can consume it at a time that's
3:48:32
can consume it at a time that's
3:48:32
can consume it at a time that's convenient
3:48:33
convenient
3:48:33
convenient to them
3:48:35
to them
3:48:35
to them um but obviously for anyone who is
3:48:37
um but obviously for anyone who is
3:48:37
um but obviously for anyone who is interested in a particular topic if they
3:48:39
interested in a particular topic if they
3:48:39
interested in a particular topic if they hop in there today tomorrow then they
3:48:41
hop in there today tomorrow then they
3:48:41
hop in there today tomorrow then they can ask questions on the Discord and
3:48:43
can ask questions on the Discord and
3:48:43
can ask questions on the Discord and actually it's a little bit easier to
3:48:44
actually it's a little bit easier to
3:48:45
actually it's a little bit easier to reach the speakers unless those speakers
3:48:47
reach the speakers unless those speakers
3:48:47
reach the speakers unless those speakers are a cubecon a few people are
3:48:49
are a cubecon a few people are
3:48:49
are a cubecon a few people are distracted with kubecon
3:48:51
distracted with kubecon
3:48:51
distracted with kubecon um there are other conferences going on
3:48:53
um there are other conferences going on
3:48:53
um there are other conferences going on there are other conferences I know
3:48:55
there are other conferences I know
3:48:55
there are other conferences I know there's a Microsoft MVP Summit too for
3:48:57
there's a Microsoft MVP Summit too for
3:48:57
there's a Microsoft MVP Summit too for people who are involved in that
3:48:59
people who are involved in that
3:48:59
people who are involved in that um so we have a couple of promo videos
3:49:01
um so we have a couple of promo videos
3:49:01
um so we have a couple of promo videos and I wanted to play that that first one
3:49:05
and I wanted to play that that first one
3:49:05
and I wanted to play that that first one that is all about postgres performance
3:49:07
that is all about postgres performance
3:49:07
that is all about postgres performance and security that goes through super
3:49:09
and security that goes through super
3:49:09
and security that goes through super super fast like 10 seconds each for some
3:49:11
super fast like 10 seconds each for some
3:49:12
super fast like 10 seconds each for some of these on-demand talks um does that
3:49:13
of these on-demand talks um does that
3:49:13
of these on-demand talks um does that sound good that sounds awesome okay
3:49:15
sound good that sounds awesome okay
3:49:15
sound good that sounds awesome okay let's go
3:49:21
thank you
3:49:24
thank you
3:49:24
thank you but in fact I'm going to show you
3:49:26
but in fact I'm going to show you
3:49:26
but in fact I'm going to show you different hopefully exciting ways that
3:49:29
different hopefully exciting ways that
3:49:29
different hopefully exciting ways that you don't know about to index uuids and
3:49:33
you don't know about to index uuids and
3:49:33
you don't know about to index uuids and postgresql
3:49:34
postgresql
3:49:34
postgresql so if you are a traveling postgres
3:49:38
so if you are a traveling postgres
3:49:38
so if you are a traveling postgres consultant like me you will see that
3:49:40
consultant like me you will see that
3:49:40
consultant like me you will see that often
3:49:42
often
3:49:42
often um
3:49:42
um
3:49:42
um a lot of performance can be gained by
3:49:45
a lot of performance can be gained by
3:49:45
a lot of performance can be gained by outsmarting people right
3:49:48
outsmarting people right
3:49:48
outsmarting people right today I'm going to be talking about
3:49:49
today I'm going to be talking about
3:49:49
today I'm going to be talking about postgres tableblo but the overall idea
3:49:51
postgres tableblo but the overall idea
3:49:51
postgres tableblo but the overall idea of this talk is discussing the
3:49:54
of this talk is discussing the
3:49:54
of this talk is discussing the transition from doing postgres at a
3:49:56
transition from doing postgres at a
3:49:56
transition from doing postgres at a smaller scale to post scale
3:49:59
smaller scale to post scale
3:49:59
smaller scale to post scale I'm going to talk to you today about
3:50:00
I'm going to talk to you today about
3:50:00
I'm going to talk to you today about postgresql privileges roles and security
3:50:03
postgresql privileges roles and security
3:50:03
postgresql privileges roles and security and how you can take better advantage of
3:50:06
and how you can take better advantage of
3:50:06
and how you can take better advantage of them within your postgres installation
3:50:08
them within your postgres installation
3:50:08
them within your postgres installation the thing I wanted to talk about explain
3:50:10
the thing I wanted to talk about explain
3:50:10
the thing I wanted to talk about explain the concepts behind tuning high right
3:50:13
the concepts behind tuning high right
3:50:13
the concepts behind tuning high right workloads for postgres and then dive
3:50:16
workloads for postgres and then dive
3:50:16
workloads for postgres and then dive deeper into the specific configurations
3:50:19
deeper into the specific configurations
3:50:19
deeper into the specific configurations and devops data technical lead of jfrog
3:50:22
and devops data technical lead of jfrog
3:50:22
and devops data technical lead of jfrog and I want to share my lecture about
3:50:25
and I want to share my lecture about
3:50:25
and I want to share my lecture about troubleshooting High CPU utilization for
3:50:28
troubleshooting High CPU utilization for
3:50:28
troubleshooting High CPU utilization for postgres database
3:50:30
postgres database
3:50:30
postgres database we will talk about
3:50:31
we will talk about
3:50:31
we will talk about external attacks about authentication
3:50:34
external attacks about authentication
3:50:34
external attacks about authentication security about data protection and also
3:50:38
security about data protection and also
3:50:38
security about data protection and also I will share some trips and my
3:50:41
I will share some trips and my
3:50:41
I will share some trips and my recommendations with you
3:50:43
recommendations with you
3:50:43
recommendations with you [Music]
3:50:45
[Music]
3:50:45
[Music] foreign
3:50:52
I got a chance to look at hockey slides
3:50:54
I got a chance to look at hockey slides
3:50:54
I got a chance to look at hockey slides a little bit and just started like
3:50:56
a little bit and just started like
3:50:56
a little bit and just started like peeking at that uh and that will be one
3:50:58
peeking at that uh and that will be one
3:50:58
peeking at that uh and that will be one definitely I'm watching tonight because
3:50:59
definitely I'm watching tonight because
3:50:59
definitely I'm watching tonight because he he definitely goes into things like I
3:51:02
he he definitely goes into things like I
3:51:02
he he definitely goes into things like I felt like I've dealt with uuids a lot in
3:51:04
felt like I've dealt with uuids a lot in
3:51:04
felt like I've dealt with uuids a lot in postgres he does talk about some stuff
3:51:06
postgres he does talk about some stuff
3:51:06
postgres he does talk about some stuff in there that I don't think I've seen
3:51:07
in there that I don't think I've seen
3:51:07
in there that I don't think I've seen anyone else actually talk about before
3:51:09
anyone else actually talk about before
3:51:09
anyone else actually talk about before so I maybe I missed it but I'm super
3:51:12
so I maybe I missed it but I'm super
3:51:12
so I maybe I missed it but I'm super excited on that one uh and I would all
3:51:14
excited on that one uh and I would all
3:51:14
excited on that one uh and I would all say like Ryan I know has been doing a
3:51:16
say like Ryan I know has been doing a
3:51:16
say like Ryan I know has been doing a lot of advocacy work around roles and
3:51:18
lot of advocacy work around roles and
3:51:18
lot of advocacy work around roles and grants so I'm pretty excited to check
3:51:19
grants so I'm pretty excited to check
3:51:19
grants so I'm pretty excited to check that one out as well uh and I mean they
3:51:21
that one out as well uh and I mean they
3:51:21
that one out as well uh and I mean they all I think look interesting at least on
3:51:23
all I think look interesting at least on
3:51:23
all I think look interesting at least on first glance so that's that's pretty
3:51:25
first glance so that's that's pretty
3:51:25
first glance so that's that's pretty good yeah I think Ryan gave a talk at
3:51:27
good yeah I think Ryan gave a talk at
3:51:27
good yeah I think Ryan gave a talk at scale and you were there as well I don't
3:51:29
scale and you were there as well I don't
3:51:29
scale and you were there as well I don't know if you saw saw that one and then
3:51:31
know if you saw saw that one and then
3:51:31
know if you saw saw that one and then Chelsea Dole I saw her give a lightning
3:51:34
Chelsea Dole I saw her give a lightning
3:51:34
Chelsea Dole I saw her give a lightning talk at pgconf EU in Berlin a couple of
3:51:37
talk at pgconf EU in Berlin a couple of
3:51:37
talk at pgconf EU in Berlin a couple of months ago and
3:51:39
months ago and
3:51:39
months ago and um yeah she was she was fabulous I went
3:51:42
um yeah she was she was fabulous I went
3:51:42
um yeah she was she was fabulous I went up introduced myself sit told her all
3:51:44
up introduced myself sit told her all
3:51:44
up introduced myself sit told her all about satiscon and said please submit a
3:51:46
about satiscon and said please submit a
3:51:46
about satiscon and said please submit a talk proposal
3:51:47
talk proposal
3:51:47
talk proposal um although I mean the talks that get
3:51:50
um although I mean the talks that get
3:51:50
um although I mean the talks that get accepted in the status con is pretty
3:51:52
accepted in the status con is pretty
3:51:52
accepted in the status con is pretty competitive I think we have like a 27
3:51:55
competitive I think we have like a 27
3:51:55
competitive I think we have like a 27 acceptance rate
3:51:57
acceptance rate
3:51:57
acceptance rate um so 32 invited Keynotes obviously but
3:52:00
um so 32 invited Keynotes obviously but
3:52:00
um so 32 invited Keynotes obviously but 35 talks accepted so you can do the
3:52:02
35 talks accepted so you can do the
3:52:02
35 talks accepted so you can do the quick math in your head and figure out
3:52:04
quick math in your head and figure out
3:52:04
quick math in your head and figure out how many submissions there were there
3:52:06
how many submissions there were there
3:52:06
how many submissions there were there were a lot of phenomenal submissions and
3:52:08
were a lot of phenomenal submissions and
3:52:08
were a lot of phenomenal submissions and it was really hard for the the talks
3:52:10
it was really hard for the the talks
3:52:10
it was really hard for the the talks election team
3:52:12
election team
3:52:12
election team um and in fact I think I have a slide
3:52:14
um and in fact I think I have a slide
3:52:14
um and in fact I think I have a slide here
3:52:15
here
3:52:15
here um if I click ahead yeah there it was
3:52:17
um if I click ahead yeah there it was
3:52:17
um if I click ahead yeah there it was just to say thank you to everybody who
3:52:19
just to say thank you to everybody who
3:52:19
just to say thank you to everybody who was on the the talk selection team
3:52:22
was on the the talk selection team
3:52:22
was on the the talk selection team um Aaron whistling Alicia Marco and
3:52:25
um Aaron whistling Alicia Marco and
3:52:25
um Aaron whistling Alicia Marco and Charles
3:52:26
Charles
3:52:26
Charles um spent a lot of time reviewing all
3:52:28
um spent a lot of time reviewing all
3:52:28
um spent a lot of time reviewing all those proposals and this conference you
3:52:30
those proposals and this conference you
3:52:30
those proposals and this conference you know obviously I say every great event
3:52:32
know obviously I say every great event
3:52:32
know obviously I say every great event starts with great speakers but um it the
3:52:35
starts with great speakers but um it the
3:52:35
starts with great speakers but um it the talk selection team does a lot of work
3:52:36
talk selection team does a lot of work
3:52:36
talk selection team does a lot of work to get us there
3:52:38
to get us there
3:52:38
to get us there absolutely and uh you know again they
3:52:41
absolutely and uh you know again they
3:52:41
absolutely and uh you know again they made sure not to invite me back so I
3:52:42
made sure not to invite me back so I
3:52:42
made sure not to invite me back so I think the level this year has gone up
3:52:44
think the level this year has gone up
3:52:44
think the level this year has gone up even more considerably uh so that's ouch
3:52:47
even more considerably uh so that's ouch
3:52:47
even more considerably uh so that's ouch don't be self-def
3:52:49
ER those of you who care about Swag um
3:52:52
ER those of you who care about Swag um
3:52:52
ER those of you who care about Swag um the code is on screen right now as well
3:52:54
the code is on screen right now as well
3:52:54
the code is on screen right now as well as the URL there are 75 of these being
3:52:56
as the URL there are 75 of these being
3:52:57
as the URL there are 75 of these being given away in this live stream time
3:52:58
given away in this live stream time
3:52:58
given away in this live stream time period so if you haven't gone to do it
3:53:00
period so if you haven't gone to do it
3:53:00
period so if you haven't gone to do it yet
3:53:01
yet
3:53:01
yet um the activity book is fun not only for
3:53:04
um the activity book is fun not only for
3:53:04
um the activity book is fun not only for kids but I remember when I first met
3:53:06
kids but I remember when I first met
3:53:06
kids but I remember when I first met Melanie plagueman
3:53:08
Melanie plagueman
3:53:08
Melanie plagueman um it was because she had gotten the
3:53:09
um it was because she had gotten the
3:53:09
um it was because she had gotten the previous version of the activity book at
3:53:11
previous version of the activity book at
3:53:11
previous version of the activity book at a postgres conference and she was
3:53:13
a postgres conference and she was
3:53:13
a postgres conference and she was enjoying um coloring in it so uh uh yeah
3:53:18
enjoying um coloring in it so uh uh yeah
3:53:18
enjoying um coloring in it so uh uh yeah um nothing better for when you are
3:53:19
um nothing better for when you are
3:53:19
um nothing better for when you are waiting on our index rebuild uh you know
3:53:22
waiting on our index rebuild uh you know
3:53:22
waiting on our index rebuild uh you know than to be able to whip out the activity
3:53:23
than to be able to whip out the activity
3:53:23
than to be able to whip out the activity book and uh you know pass some time that
3:53:26
book and uh you know pass some time that
3:53:26
book and uh you know pass some time that way and socks are always cool and
3:53:28
way and socks are always cool and
3:53:28
way and socks are always cool and there's the sticker packs too for those
3:53:29
there's the sticker packs too for those
3:53:29
there's the sticker packs too for those of you who use stickers either on your
3:53:31
of you who use stickers either on your
3:53:31
of you who use stickers either on your laptops like me or on your luggage I
3:53:33
laptops like me or on your luggage I
3:53:33
laptops like me or on your luggage I feel like I'm in an infomercial but I
3:53:35
feel like I'm in an infomercial but I
3:53:35
feel like I'm in an infomercial but I know how much people care about Swag
3:53:37
know how much people care about Swag
3:53:37
know how much people care about Swag it's like it's a thing and so speakers
3:53:39
it's like it's a thing and so speakers
3:53:39
it's like it's a thing and so speakers yeah they do okay
3:53:42
yeah they do okay
3:53:42
yeah they do okay um we have another quick video trailer
3:53:44
um we have another quick video trailer
3:53:44
um we have another quick video trailer this one is focused more on the two
3:53:46
this one is focused more on the two
3:53:46
this one is focused more on the two Azure database services for postgres
3:53:49
Azure database services for postgres
3:53:49
Azure database services for postgres Azure Cosmos DB for postgres other what
3:53:53
Azure Cosmos DB for postgres other what
3:53:53
Azure Cosmos DB for postgres other what I happen to call cytus on Azure because
3:53:55
I happen to call cytus on Azure because
3:53:55
I happen to call cytus on Azure because I'm such a sightest open source person
3:53:57
I'm such a sightest open source person
3:53:57
I'm such a sightest open source person and uh and then also Azure database for
3:54:01
and uh and then also Azure database for
3:54:01
and uh and then also Azure database for a postgres flex server so let's roll
3:54:03
a postgres flex server so let's roll
3:54:03
a postgres flex server so let's roll this this promo
3:54:09
it's number two
3:54:11
it's number two
3:54:11
it's number two unless you want me to do
3:54:15
[Music]
3:54:17
[Music]
3:54:17
[Music] hey Lucas here and I'll present to you
3:54:20
hey Lucas here and I'll present to you
3:54:20
hey Lucas here and I'll present to you how I alto's killed a cosmos DB for
3:54:22
how I alto's killed a cosmos DB for
3:54:22
how I alto's killed a cosmos DB for postgres cluster using cytos of course
3:54:25
postgres cluster using cytos of course
3:54:25
postgres cluster using cytos of course rafana and Azure Surplus
3:54:28
rafana and Azure Surplus
3:54:28
rafana and Azure Surplus this talk is for multi-ten and SAS
3:54:30
this talk is for multi-ten and SAS
3:54:30
this talk is for multi-ten and SAS companies who are growing fast that is
3:54:32
companies who are growing fast that is
3:54:33
companies who are growing fast that is their onboarding thousands of customers
3:54:34
their onboarding thousands of customers
3:54:34
their onboarding thousands of customers and they are expecting or already
3:54:36
and they are expecting or already
3:54:36
and they are expecting or already running into scalability issues with the
3:54:39
running into scalability issues with the
3:54:39
running into scalability issues with the database and want to scale out with
3:54:41
database and want to scale out with
3:54:41
database and want to scale out with distributed postgresql
3:54:43
distributed postgresql
3:54:43
distributed postgresql so today we are going to talk about
3:54:47
so today we are going to talk about
3:54:47
so today we are going to talk about multi-tenant software as a service
3:54:50
multi-tenant software as a service
3:54:50
multi-tenant software as a service applications on Azure customers DB for
3:54:53
applications on Azure customers DB for
3:54:53
applications on Azure customers DB for personal SQL
3:54:54
personal SQL
3:54:54
personal SQL it's built upon like the community
3:54:57
it's built upon like the community
3:54:57
it's built upon like the community version of postgres so the open source
3:54:59
version of postgres so the open source
3:54:59
version of postgres so the open source postgres and ads also
3:55:02
postgres and ads also
3:55:02
postgres and ads also um cytos extension which turns a
3:55:05
um cytos extension which turns a
3:55:05
um cytos extension which turns a postgres into distributed database
3:55:09
postgres into distributed database
3:55:09
postgres into distributed database in this session I would like to talk
3:55:12
in this session I would like to talk
3:55:12
in this session I would like to talk about partitioning in postgresql and how
3:55:14
about partitioning in postgresql and how
3:55:14
about partitioning in postgresql and how it is similar and different than Oracle
3:55:18
it is similar and different than Oracle
3:55:18
it is similar and different than Oracle so today's focus of the talk is to
3:55:21
so today's focus of the talk is to
3:55:21
so today's focus of the talk is to discuss some tools recommendations and
3:55:23
discuss some tools recommendations and
3:55:23
discuss some tools recommendations and best practices of how to reduce cost on
3:55:25
best practices of how to reduce cost on
3:55:25
best practices of how to reduce cost on Azure database Focus flexible server my
3:55:28
Azure database Focus flexible server my
3:55:28
Azure database Focus flexible server my talk is about Asia ID authentication
3:55:30
talk is about Asia ID authentication
3:55:30
talk is about Asia ID authentication with flexible servers if you use
3:55:33
with flexible servers if you use
3:55:33
with flexible servers if you use flexible servers or plan to try it out
3:55:36
flexible servers or plan to try it out
3:55:36
flexible servers or plan to try it out this feature will help make your
3:55:38
this feature will help make your
3:55:38
this feature will help make your application connections more secure it
3:55:40
application connections more secure it
3:55:40
application connections more secure it will save your time managing uh
3:55:42
will save your time managing uh
3:55:42
will save your time managing uh credentials and roles so please check it
3:55:45
credentials and roles so please check it
3:55:45
credentials and roles so please check it out
3:55:45
out
3:55:45
out [Music]
3:55:50
[Music]
3:55:50
[Music] awesome uh I'm I'm not gonna be like
3:55:53
awesome uh I'm I'm not gonna be like
3:55:53
awesome uh I'm I'm not gonna be like every one of these is awesome so but I
3:55:55
every one of these is awesome so but I
3:55:55
every one of these is awesome so but I will pick one in particular that I have
3:55:56
will pick one in particular that I have
3:55:56
will pick one in particular that I have to say so I've just I've seen a lot of
3:55:58
to say so I've just I've seen a lot of
3:55:58
to say so I've just I've seen a lot of chatter about multi-tenancy uh
3:56:01
chatter about multi-tenancy uh
3:56:01
chatter about multi-tenancy uh deployments in postgres lately uh Super
3:56:03
deployments in postgres lately uh Super
3:56:03
deployments in postgres lately uh Super Hot Topic uh and I it's like it's come
3:56:05
Hot Topic uh and I it's like it's come
3:56:05
Hot Topic uh and I it's like it's come back into fashion I guess
3:56:07
back into fashion I guess
3:56:07
back into fashion I guess um so that one I'm super interested to
3:56:09
um so that one I'm super interested to
3:56:09
um so that one I'm super interested to see because I to me I don't know if
3:56:11
see because I to me I don't know if
3:56:11
see because I to me I don't know if people always think of cytus as a
3:56:12
people always think of cytus as a
3:56:12
people always think of cytus as a multi-tenancy solution
3:56:14
multi-tenancy solution
3:56:14
multi-tenancy solution um but clearly it can be used for that
3:56:16
um but clearly it can be used for that
3:56:16
um but clearly it can be used for that and so that means that means I wasn't
3:56:18
and so that means that means I wasn't
3:56:18
and so that means that means I wasn't doing my job I mean before the
3:56:20
doing my job I mean before the
3:56:20
doing my job I mean before the acquisition
3:56:21
acquisition
3:56:21
acquisition um when we came out with the worry-free
3:56:23
um when we came out with the worry-free
3:56:23
um when we came out with the worry-free postgres tagline because we decided that
3:56:25
postgres tagline because we decided that
3:56:25
postgres tagline because we decided that most people only had 30 000 days in
3:56:28
most people only had 30 000 days in
3:56:28
most people only had 30 000 days in their life and time is our most precious
3:56:30
their life and time is our most precious
3:56:30
their life and time is our most precious Resource as developers as Engineers as
3:56:33
Resource as developers as Engineers as
3:56:33
Resource as developers as Engineers as people who manage databases and so we
3:56:36
people who manage databases and so we
3:56:36
people who manage databases and so we were like you know what people don't
3:56:37
were like you know what people don't
3:56:37
were like you know what people don't want to wake up at three in the morning
3:56:38
want to wake up at three in the morning
3:56:38
want to wake up at three in the morning to deal with postgres database issues
3:56:40
to deal with postgres database issues
3:56:40
to deal with postgres database issues and so let's focus on worry-free
3:56:43
and so let's focus on worry-free
3:56:43
and so let's focus on worry-free postgres but the other alternative at
3:56:45
postgres but the other alternative at
3:56:45
postgres but the other alternative at the time was to kind of really focus on
3:56:47
the time was to kind of really focus on
3:56:47
the time was to kind of really focus on driving home the point that cytus was a
3:56:50
driving home the point that cytus was a
3:56:50
driving home the point that cytus was a multi-tenant data space so maybe you
3:56:53
multi-tenant data space so maybe you
3:56:53
multi-tenant data space so maybe you don't connect those dots because we've
3:56:55
don't connect those dots because we've
3:56:55
don't connect those dots because we've focused on that worry-free angle but for
3:56:58
focused on that worry-free angle but for
3:56:58
focused on that worry-free angle but for me yeah so the worry free was definitely
3:57:00
me yeah so the worry free was definitely
3:57:00
me yeah so the worry free was definitely a big part but also I think using it for
3:57:02
a big part but also I think using it for
3:57:02
a big part but also I think using it for like uh statistics monitoring olap like
3:57:05
like uh statistics monitoring olap like
3:57:05
like uh statistics monitoring olap like that type of stuff like we saw in the
3:57:07
that type of stuff like we saw in the
3:57:07
that type of stuff like we saw in the earlier talk with Json and you're right
3:57:09
earlier talk with Json and you're right
3:57:09
earlier talk with Json and you're right and doing that monitoring piece to me I
3:57:11
and doing that monitoring piece to me I
3:57:11
and doing that monitoring piece to me I think a lot of people connected with
3:57:12
think a lot of people connected with
3:57:12
think a lot of people connected with that type of a workload
3:57:14
that type of a workload
3:57:14
that type of a workload um you know and so that that's the
3:57:16
um you know and so that that's the
3:57:16
um you know and so that that's the reason why I say that I don't know if
3:57:17
reason why I say that I don't know if
3:57:17
reason why I say that I don't know if everyone connects it with both sides of
3:57:18
everyone connects it with both sides of
3:57:18
everyone connects it with both sides of that it's good for both of those really
3:57:20
that it's good for both of those really
3:57:20
that it's good for both of those really yeah oh and and it is good for both and
3:57:23
yeah oh and and it is good for both and
3:57:23
yeah oh and and it is good for both and we have tried to serve both of those use
3:57:25
we have tried to serve both of those use
3:57:25
we have tried to serve both of those use cases both multi-tenant SAS and
3:57:26
cases both multi-tenant SAS and
3:57:26
cases both multi-tenant SAS and real-time analytics but I'm not here to
3:57:29
real-time analytics but I'm not here to
3:57:29
real-time analytics but I'm not here to to promo the project right now although
3:57:31
to promo the project right now although
3:57:31
to promo the project right now although I did just pop up the GitHub the GitHub
3:57:33
I did just pop up the GitHub the GitHub
3:57:33
I did just pop up the GitHub the GitHub URL on screen
3:57:35
URL on screen
3:57:35
URL on screen um okay so let me see I've got a promo
3:57:40
um okay so let me see I've got a promo
3:57:40
um okay so let me see I've got a promo while you're prepping the video uh that
3:57:42
while you're prepping the video uh that
3:57:42
while you're prepping the video uh that you know go on GitHub uh it's all open
3:57:44
you know go on GitHub uh it's all open
3:57:44
you know go on GitHub uh it's all open source again this is one of the honestly
3:57:46
source again this is one of the honestly
3:57:46
source again this is one of the honestly one of the reasons I'm here it's all
3:57:48
one of the reasons I'm here it's all
3:57:48
one of the reasons I'm here it's all open source it's on GitHub uh you know
3:57:51
open source it's on GitHub uh you know
3:57:51
open source it's on GitHub uh you know star and and watch the repo all that
3:57:54
star and and watch the repo all that
3:57:54
star and and watch the repo all that good stuff uh would recommend anyone
3:57:56
good stuff uh would recommend anyone
3:57:56
good stuff uh would recommend anyone who's interested in cytus uh there's a
3:57:58
who's interested in cytus uh there's a
3:57:58
who's interested in cytus uh there's a lot of interesting stuff that goes on
3:57:59
lot of interesting stuff that goes on
3:57:59
lot of interesting stuff that goes on over here so you know well you've been a
3:58:01
over here so you know well you've been a
3:58:01
over here so you know well you've been a supporter of the project from far away
3:58:03
supporter of the project from far away
3:58:03
supporter of the project from far away for a long time and I I really
3:58:05
for a long time and I I really
3:58:05
for a long time and I I really appreciate it because a rising tide
3:58:07
appreciate it because a rising tide
3:58:07
appreciate it because a rising tide floats all boats like um I mean I
3:58:09
floats all boats like um I mean I
3:58:09
floats all boats like um I mean I obviously I'm a big fan of other
3:58:10
obviously I'm a big fan of other
3:58:10
obviously I'm a big fan of other projects I love the fact that Lucas did
3:58:13
projects I love the fact that Lucas did
3:58:13
projects I love the fact that Lucas did a demo with grafana which I'm a huge fan
3:58:15
a demo with grafana which I'm a huge fan
3:58:15
a demo with grafana which I'm a huge fan of and and I think it's good when we we
3:58:18
of and and I think it's good when we we
3:58:18
of and and I think it's good when we we support each other or recommend the
3:58:19
support each other or recommend the
3:58:19
support each other or recommend the whole word of mouth thing is useful
3:58:22
whole word of mouth thing is useful
3:58:22
whole word of mouth thing is useful um in the developer world we have
3:58:24
um in the developer world we have
3:58:24
um in the developer world we have um a couple of other I know we're tied
3:58:26
um a couple of other I know we're tied
3:58:26
um a couple of other I know we're tied on time because we have to end in just a
3:58:28
on time because we have to end in just a
3:58:28
on time because we have to end in just a couple minutes and we have two more
3:58:29
couple minutes and we have two more
3:58:29
couple minutes and we have two more videos we want to roll so let's go ahead
3:58:31
videos we want to roll so let's go ahead
3:58:31
videos we want to roll so let's go ahead and roll video number three please
3:58:37
[Music]
3:58:39
[Music]
3:58:39
[Music] so this talk we're gonna discuss a
3:58:42
so this talk we're gonna discuss a
3:58:42
so this talk we're gonna discuss a little bit why you want to run a
3:58:43
little bit why you want to run a
3:58:43
little bit why you want to run a database on kubernetes how this was done
3:58:46
database on kubernetes how this was done
3:58:46
database on kubernetes how this was done to run cytos to create cytos clusters
3:58:49
to run cytos to create cytos clusters
3:58:49
to run cytos to create cytos clusters before patroni 3.0 appeared and the most
3:58:52
before patroni 3.0 appeared and the most
3:58:52
before patroni 3.0 appeared and the most important thing how you can run
3:58:53
important thing how you can run
3:58:53
important thing how you can run kubernetes on kubernetes today with open
3:58:57
kubernetes on kubernetes today with open
3:58:57
kubernetes on kubernetes today with open source software in a very very easy
3:58:59
source software in a very very easy
3:58:59
source software in a very very easy manner in my opinion cyber security
3:59:02
manner in my opinion cyber security
3:59:02
manner in my opinion cyber security applications have a certain unique set
3:59:05
applications have a certain unique set
3:59:05
applications have a certain unique set of features and requirements that are
3:59:08
of features and requirements that are
3:59:08
of features and requirements that are different from other applications I also
3:59:10
different from other applications I also
3:59:10
different from other applications I also think that scientist cytus database is a
3:59:14
think that scientist cytus database is a
3:59:14
think that scientist cytus database is a very uniquely positioned to solve many
3:59:17
very uniquely positioned to solve many
3:59:17
very uniquely positioned to solve many of the challenges that cyber security
3:59:19
of the challenges that cyber security
3:59:19
of the challenges that cyber security applications present I'm a principal
3:59:21
applications present I'm a principal
3:59:21
applications present I'm a principal engineer working on a platform here at
3:59:23
engineer working on a platform here at
3:59:23
engineer working on a platform here at jellyfish I've been here since the early
3:59:25
jellyfish I've been here since the early
3:59:25
jellyfish I've been here since the early days and through a lot of our of our
3:59:27
days and through a lot of our of our
3:59:27
days and through a lot of our of our growth and our infra choices including
3:59:30
growth and our infra choices including
3:59:30
growth and our infra choices including the the move to a sharded database that
3:59:32
the the move to a sharded database that
3:59:32
the the move to a sharded database that we're going to talk with you here today
3:59:35
we're going to talk with you here today
3:59:35
we're going to talk with you here today about
3:59:39
ahead of some of the gotchas that come
3:59:41
ahead of some of the gotchas that come
3:59:41
ahead of some of the gotchas that come up when trying to move to a self-hosted
3:59:43
up when trying to move to a self-hosted
3:59:43
up when trying to move to a self-hosted sinus implementation we went through
3:59:46
sinus implementation we went through
3:59:46
sinus implementation we went through this journey and we'd like to share some
3:59:50
this journey and we'd like to share some
3:59:50
this journey and we'd like to share some of the interesting technical challenges
3:59:51
of the interesting technical challenges
3:59:51
of the interesting technical challenges that we had to solve
3:59:53
that we had to solve
3:59:53
that we had to solve because I I frequently get questions
3:59:55
because I I frequently get questions
3:59:55
because I I frequently get questions about foreign key support insighters in
3:59:58
about foreign key support insighters in
3:59:58
about foreign key support insighters in this talk I want to clarify some of the
4:00:00
this talk I want to clarify some of the
4:00:00
this talk I want to clarify some of the concepts regarding scientists
4:00:03
concepts regarding scientists
4:00:03
concepts regarding scientists hello I'm here today to share with you
4:00:05
hello I'm here today to share with you
4:00:05
hello I'm here today to share with you some lessons and observations from
4:00:07
some lessons and observations from
4:00:07
some lessons and observations from safety cultures migration from a managed
4:00:09
safety cultures migration from a managed
4:00:09
safety cultures migration from a managed postgresql service to our own
4:00:10
postgresql service to our own
4:00:10
postgresql service to our own self-hosted process
4:00:11
self-hosted process
4:00:11
self-hosted process [Music]
4:00:16
[Music]
4:00:16
[Music] also another fine slate of talks uh I
4:00:19
also another fine slate of talks uh I
4:00:19
also another fine slate of talks uh I especially like the one moving to the
4:00:22
especially like the one moving to the
4:00:22
especially like the one moving to the self-hosted uh from the manage that's
4:00:25
self-hosted uh from the manage that's
4:00:25
self-hosted uh from the manage that's really complicated and uh I think we'll
4:00:27
really complicated and uh I think we'll
4:00:27
really complicated and uh I think we'll explain to a lot of people what they're
4:00:29
explain to a lot of people what they're
4:00:29
explain to a lot of people what they're getting into
4:00:30
getting into
4:00:30
getting into so yeah it was a pleasure pleasure for
4:00:33
so yeah it was a pleasure pleasure for
4:00:33
so yeah it was a pleasure pleasure for me to meet
4:00:34
me to meet
4:00:34
me to meet and Delaney McKenzie from jellyfish as
4:00:37
and Delaney McKenzie from jellyfish as
4:00:37
and Delaney McKenzie from jellyfish as they told that story although also Paul
4:00:39
they told that story although also Paul
4:00:39
they told that story although also Paul diadny told a similar story
4:00:41
diadny told a similar story
4:00:42
diadny told a similar story um about moving to a self-hosted sinus
4:00:45
um about moving to a self-hosted sinus
4:00:45
um about moving to a self-hosted sinus environment so
4:00:46
environment so
4:00:46
environment so um obviously you know I work at
4:00:48
um obviously you know I work at
4:00:48
um obviously you know I work at Microsoft and uh for people who want
4:00:51
Microsoft and uh for people who want
4:00:51
Microsoft and uh for people who want that managed service offering we we
4:00:53
that managed service offering we we
4:00:53
that managed service offering we we offer that but it's I love I love seeing
4:00:56
offer that but it's I love I love seeing
4:00:56
offer that but it's I love I love seeing people be successful with status open
4:00:58
people be successful with status open
4:00:58
people be successful with status open source too
4:00:59
source too
4:01:00
source too um okay so I think we have let's see
4:01:03
um okay so I think we have let's see
4:01:03
um okay so I think we have let's see another video here that some more
4:01:07
another video here that some more
4:01:07
another video here that some more postgres Community talks so let's take a
4:01:09
postgres Community talks so let's take a
4:01:09
postgres Community talks so let's take a look at these
4:01:14
[Music]
4:01:16
[Music]
4:01:16
[Music] I am Dimitri fonten and I've been
4:01:19
I am Dimitri fonten and I've been
4:01:19
I am Dimitri fonten and I've been contributing to postgresql for a very
4:01:20
contributing to postgresql for a very
4:01:21
contributing to postgresql for a very long time now we're going to talk
4:01:23
long time now we're going to talk
4:01:23
long time now we're going to talk about how to copy your postgres database
4:01:26
about how to copy your postgres database
4:01:26
about how to copy your postgres database so from one server to the other one
4:01:29
so from one server to the other one
4:01:30
so from one server to the other one the concept of storing time with
4:01:33
the concept of storing time with
4:01:33
the concept of storing time with databases have been popular for very
4:01:36
databases have been popular for very
4:01:36
databases have been popular for very long time pretty much since relational
4:01:38
long time pretty much since relational
4:01:38
long time pretty much since relational databases come to life
4:01:41
databases come to life
4:01:41
databases come to life exactly the goal this talk is to talk
4:01:43
exactly the goal this talk is to talk
4:01:43
exactly the goal this talk is to talk about how postgres can give you an ideal
4:01:46
about how postgres can give you an ideal
4:01:46
about how postgres can give you an ideal platform to use artificial intelligence
4:01:49
platform to use artificial intelligence
4:01:49
platform to use artificial intelligence uh in your applications
4:01:52
uh in your applications
4:01:52
uh in your applications we'll set the scene by imagining that
4:01:54
we'll set the scene by imagining that
4:01:54
we'll set the scene by imagining that you're the DBA in an organization that
4:01:57
you're the DBA in an organization that
4:01:57
you're the DBA in an organization that needs to implement multi-tenancy
4:02:00
needs to implement multi-tenancy
4:02:00
needs to implement multi-tenancy what I'm going to talk about today is
4:02:02
what I'm going to talk about today is
4:02:02
what I'm going to talk about today is how we've taken postgresql an insane
4:02:05
how we've taken postgresql an insane
4:02:05
how we've taken postgresql an insane relational database and how we've
4:02:08
relational database and how we've
4:02:08
relational database and how we've Twisted it and molded it and created it
4:02:10
Twisted it and molded it and created it
4:02:10
Twisted it and molded it and created it what we think is very robust document
4:02:12
what we think is very robust document
4:02:12
what we think is very robust document database and event store solution
4:02:16
database and event store solution
4:02:16
database and event store solution today I will talk about in-depth Guide
4:02:18
today I will talk about in-depth Guide
4:02:18
today I will talk about in-depth Guide to the Past Chris's new CI and I will
4:02:22
to the Past Chris's new CI and I will
4:02:22
to the Past Chris's new CI and I will explain how to use that new CI
4:02:25
explain how to use that new CI
4:02:25
explain how to use that new CI in this talk we see together how to
4:02:27
in this talk we see together how to
4:02:27
in this talk we see together how to build web maps from scratch using Django
4:02:30
build web maps from scratch using Django
4:02:30
build web maps from scratch using Django and postgis if you are asking yourself
4:02:32
and postgis if you are asking yourself
4:02:32
and postgis if you are asking yourself what type of maps we can build with
4:02:35
what type of maps we can build with
4:02:35
what type of maps we can build with chiangle let's see an example right away
4:02:38
chiangle let's see an example right away
4:02:38
chiangle let's see an example right away [Music]
4:02:40
[Music]
4:02:40
[Music] foreign
4:02:46
there's a playlist well actually just go
4:02:49
there's a playlist well actually just go
4:02:49
there's a playlist well actually just go to aka.ms slash on demand and um you can
4:02:54
to aka.ms slash on demand and um you can
4:02:54
to aka.ms slash on demand and um you can get access just just go to the schedule
4:02:56
get access just just go to the schedule
4:02:56
get access just just go to the schedule page and all the on-demand talks have
4:02:58
page and all the on-demand talks have
4:02:58
page and all the on-demand talks have YouTube embeds there on that on demand
4:03:00
YouTube embeds there on that on demand
4:03:00
YouTube embeds there on that on demand Tab and so you can track down any of
4:03:03
Tab and so you can track down any of
4:03:03
Tab and so you can track down any of these and start watching
4:03:04
these and start watching
4:03:04
these and start watching so much goodness um
4:03:06
so much goodness um
4:03:06
so much goodness um I actually promised somebody I would pop
4:03:09
I actually promised somebody I would pop
4:03:09
I actually promised somebody I would pop up a slide I know we're almost out of
4:03:11
up a slide I know we're almost out of
4:03:11
up a slide I know we're almost out of time and we got to talk about what's
4:03:12
time and we got to talk about what's
4:03:12
time and we got to talk about what's next but I am just going to pop up this
4:03:14
next but I am just going to pop up this
4:03:14
next but I am just going to pop up this slide like I promised um for people who
4:03:17
slide like I promised um for people who
4:03:17
slide like I promised um for people who do want to learn more about cytus on
4:03:19
do want to learn more about cytus on
4:03:19
do want to learn more about cytus on Azure this new Azure Cosmos DB for post
4:03:22
Azure this new Azure Cosmos DB for post
4:03:22
Azure this new Azure Cosmos DB for post press training guide came out a couple
4:03:24
press training guide came out a couple
4:03:24
press training guide came out a couple of months ago I think it was maybe
4:03:26
of months ago I think it was maybe
4:03:26
of months ago I think it was maybe October uh Joe Nelson and my team was
4:03:28
October uh Joe Nelson and my team was
4:03:28
October uh Joe Nelson and my team was involved in the creation of it not on my
4:03:30
involved in the creation of it not on my
4:03:30
involved in the creation of it not on my team but on the broader team and uh and
4:03:34
team but on the broader team and uh and
4:03:34
team but on the broader team and uh and it's it's everybody's recommending it so
4:03:36
it's it's everybody's recommending it so
4:03:36
it's it's everybody's recommending it so if you're new and you want to try this
4:03:38
if you're new and you want to try this
4:03:39
if you're new and you want to try this out and get started
4:03:40
out and get started
4:03:40
out and get started um and you want to try it on Azure
4:03:42
um and you want to try it on Azure
4:03:42
um and you want to try it on Azure versus you know just downloading the
4:03:43
versus you know just downloading the
4:03:43
versus you know just downloading the open source packages this is a really
4:03:45
open source packages this is a really
4:03:45
open source packages this is a really good place to get started I thought it
4:03:48
good place to get started I thought it
4:03:48
good place to get started I thought it was going to be a free ketchup slide but
4:03:49
was going to be a free ketchup slide but
4:03:49
was going to be a free ketchup slide but I guess this is uh much more useful and
4:03:52
I guess this is uh much more useful and
4:03:52
I guess this is uh much more useful and much easier for people who want to get
4:03:54
much easier for people who want to get
4:03:54
much easier for people who want to get started and check out some of this stuff
4:03:55
started and check out some of this stuff
4:03:55
started and check out some of this stuff so free ketchup okay
4:03:59
so free ketchup okay
4:03:59
so free ketchup okay a few minutes over time I just wanted to
4:04:01
a few minutes over time I just wanted to
4:04:01
a few minutes over time I just wanted to flag a couple things the first thing is
4:04:03
flag a couple things the first thing is
4:04:03
flag a couple things the first thing is if you're going to be awake during the
4:04:04
if you're going to be awake during the
4:04:04
if you're going to be awake during the emea live stream which is you know not
4:04:07
emea live stream which is you know not
4:04:07
emea live stream which is you know not that far away now
4:04:09
that far away now
4:04:09
that far away now um it's happening on Wednesday at 9 A.M
4:04:12
um it's happening on Wednesday at 9 A.M
4:04:12
um it's happening on Wednesday at 9 A.M central European summertime you can add
4:04:14
central European summertime you can add
4:04:14
central European summertime you can add it to your calendar with that URL in the
4:04:16
it to your calendar with that URL in the
4:04:16
it to your calendar with that URL in the top left
4:04:18
top left
4:04:18
top left um
4:04:19
um
4:04:19
um let's see no is that right
4:04:22
let's see no is that right
4:04:22
let's see no is that right yes attend emea and then the link on the
4:04:26
yes attend emea and then the link on the
4:04:26
yes attend emea and then the link on the replay the America's live stream is
4:04:28
replay the America's live stream is
4:04:28
replay the America's live stream is wrong so
4:04:30
wrong so
4:04:30
wrong so um the link on the replay of the
4:04:31
um the link on the replay of the
4:04:31
um the link on the replay of the America's live stream is when I'm
4:04:33
America's live stream is when I'm
4:04:33
America's live stream is when I'm popping on the bottom of the screen now
4:04:34
popping on the bottom of the screen now
4:04:34
popping on the bottom of the screen now in a banner so uh sorry about that folks
4:04:38
in a banner so uh sorry about that folks
4:04:38
in a banner so uh sorry about that folks um and yeah the Discord Discord what
4:04:42
um and yeah the Discord Discord what
4:04:42
um and yeah the Discord Discord what does Discord really mean
4:04:46
um I think we'll have to say that
4:04:47
um I think we'll have to say that
4:04:47
um I think we'll have to say that discussion for another day okay uh treat
4:04:50
discussion for another day okay uh treat
4:04:50
discussion for another day okay uh treat thank you so much for being my co-host
4:04:52
thank you so much for being my co-host
4:04:52
thank you so much for being my co-host and being here I know you don't work for
4:04:53
and being here I know you don't work for
4:04:53
and being here I know you don't work for Microsoft you're not part of the status
4:04:55
Microsoft you're not part of the status
4:04:55
Microsoft you're not part of the status open source project but you are part of
4:04:57
open source project but you are part of
4:04:57
open source project but you are part of the postgres community and it has been a
4:04:59
the postgres community and it has been a
4:04:59
the postgres community and it has been a delight sure it's been my honor thank
4:05:01
delight sure it's been my honor thank
4:05:01
delight sure it's been my honor thank you so much for having me and thanks for
4:05:03
you so much for having me and thanks for
4:05:03
you so much for having me and thanks for everyone who joined us and I hope you
4:05:05
everyone who joined us and I hope you
4:05:05
everyone who joined us and I hope you check out the rest of the cytus contacts
4:05:07
check out the rest of the cytus contacts
4:05:07
check out the rest of the cytus contacts and uh have a