The Cloud Show with Magnus Mårtensson ft. Annie Talvasto - Ep. 19

0:00
Hello again, everyone
0:02
Welcome back to another episode of the Cloud Show. As usual, we always have a good guest
0:09
and today we have a great one. We have Anita Lasto with us today
0:12
She is a CNCF Ambassador, that means Cloud Native Computing Foundation. She is a Microsoft MVP and she is the CMO of VSHN
0:22
I do not know what that is. We'll ask her. Today, the topic is going to be Cloud Compute Hosting Options
0:29
and when to use Kubernetes. Ani is really good with Kubernetes. When to use it and when to not
0:35
when do you use a PaaS service instead. Thank you for joining us today on the Cloud Show
0:40
Hello, Ani
0:52
Hi. Glad to be here. Yeah. Good to see you. You are all over the place
0:59
You are from Finland, but you live in Switzerland and various things. What's going on
1:05
Yeah. I live in Zurich, Switzerland for two years now, so relatively fresh facing these corners
1:12
But I do speak a lot around the world. During 2022, for example
1:17
I did 30 conferences all the way from America to Australia, to Japan and whatnot
1:23
so it's a bit of travel. That's great. You keep yourself busy with CNCF
1:30
with being an MVP, being in the community, speaking all those things
1:34
but you also have a day job. What's VSHN? Yeah. It's pronounced vision
1:39
so just like all in one go. It's one of the leading companies in Switzerland for DevOps
1:46
We have both services and products around DevOps, Kubernetes, containers, OpenShift, all of these things
1:52
I am the CMO there, so I handle marketing, but I am very much a coder and technical person at heart
1:57
which I think you have to be if you're working with DevOps topics full-time
2:03
I agree with that. I know that you can deal with all the things
2:07
DevOps obviously, and plenty with Kubernetes, which is the reason actually that I wanted to talk to you
2:14
When I said I want to talk to Annie, what should we talk about? It was like Kubernetes was the first thing that came up
2:19
You're happy to go with that? Yeah, of course. Kubernetes is one of
2:23
my favorite topics in the world. Point of question to start off, of course
2:28
I see a lot of companies start to use containerized applications and setting up
2:34
clusters and running their clusters, maybe running them in the Cloud and so forth
2:37
and then finding that they are very challenged with actually maintaining that
2:42
securing it, updating it, keeping it, just keeping it, the upkeep, the operations
2:47
Is it true that a lot of companies are going to containers because they should be going to containers
2:55
not because they actually know what they're doing? I think that's a good and a challenging question
3:00
I think the answer is, well, it's yes and no at the same time
3:05
Because I think Kubernetes and containers, particularly Kubernetes in particular, and containers as well
3:10
is one of those technologies that you can essentially use them for any use case
3:15
and it wouldn't technically be wrong. But at the same time, do you need it
3:20
Might be more of a relevant question in some cases. Kubernetes, for example, really is built for scale
3:27
It's really built for all of those scaling cases. It is container orchestration because it is about
3:34
managing bunch of containers very efficiently at the same go. Then usually, the first question that you can ask yourself
3:43
do I need Kubernetes for this case, is that do you have the scale that warrants this
3:48
If you want to do a practice round, of course, go ahead, host your personal block there
3:53
Maybe wouldn't recommend it, but you can go crazy. You can do that
3:56
Yeah, exactly. But obviously, after you answer the question of, do I use Kubernetes
4:01
If the answer is yes, then you have to go to the question of, how do I use Kubernetes
4:06
Which I think it's even more of a bigger question in some cases because it is a bit of a tricky thing to do
4:12
tricky to use, and there's so many things that you can do with it
4:16
A lot of questions there. I know. With something as technically capable and rather advanced
4:25
such as Kubernetes, running your own clusters and all those things, with that comes quite a bit of responsibility
4:31
You want a lot of power, then you have a lot of responsibility, as the expression goes
4:37
I've been saying that for the application developer who builds an application and hosts it somewhere
4:46
Kubernetes and containers and all that good stuff, it's like a platform service for the application developer
4:53
They just put their application somewhere and it runs. But that is still not really the case for
4:59
the IT operations team that take care of that environment. It's more than that for them
5:05
It's more of an, for me, that becomes more of an infrastructure service
5:10
We're taking care of a cluster. Does that make sense? It does make sense in a lot of ways
5:16
I think particularly it's interesting that you bring up the platform word or concept as a whole
5:22
because that is very much a trend word in the world of Cloud Native and Kubernetes
5:28
Because it is actually, I think, it's not even just a trend word
5:33
it is the next iteration of how we think about how you
5:37
enroll your company or how you use Kubernetes in an efficient way, for example
5:41
The platform engineering as a whole, and thinking about your Kubernetes infrastructure
5:46
or your DevOps infrastructure and all of these things as a internal development or developer platform
5:53
is taking the world by the storm. It is still relatively not early phases per se
5:59
There's a lot of material, a lot of discussion around this. But let's say that I think the most traditional companies
6:05
might not be thinking in these terms yet as of right now
6:09
So I think that is a very good take, in my opinion as well, because you're thinking about it holistically
6:14
and you're thinking about, okay, we have these people who are managing this and that's really important
6:18
But then for the end user, which would be the developers in the company, for example
6:22
it should be as easy to use and no brainer as possible
6:27
Like any platform. Yeah. Because I work with some public sector customers right now
6:33
and they come from a non-Cloud world. And that world is and has always been in a place where
6:39
oh, we're going to run a new application, we have to order a new server, right
6:44
A virtual machine in these days, but sometimes you actually have to get new hardware
6:48
because it's specific hardware and so on. So that's sort of that mindset
6:52
And on the other end of the scale is the mindset that I don't care where it runs, just run the thing
6:58
I'm expecting it to run, I'm expecting it to have a certain capacity
7:02
and that's just dealt with, thank you very much. So on that spectrum
7:08
where Kubernetes, is that in the middle? Is it a little bit of both because you manage more
7:15
but it's still a platform service and making weird gestures. It's not fun, little sock puppets
7:24
Yeah, and I think that's an interesting question. And to be frank, I think it's in the way that, okay
7:30
Kubernetes itself is very tricky to use and to expect your end users always use Kubernetes as it is
7:36
it's just not going to happen. I nowadays, in addition to the traditional
7:41
software engineering DevOps and whatnot, I talk a lot about MLOps as well
7:45
And in particularly in those kinds of cases, expecting for data scientists, for example
7:51
who are developing AI models to have knowledge of Kubernetes they probably need some, but you have to pick your battles
7:59
and so forth, so you have to choose it. But at the same time for a fully fledged
8:07
like internal developer platform, development platform, cloud native platform, or one of these
8:12
you would usually have a bit more than Kubernetes as well in the mix. So you would have Kubernetes likely as the base of it
8:18
in some ways, and then you would have a lot of the, for example, CNCF add-ons from there
8:23
So you would likely use, if you need auto scaling, you would use KEDA
8:28
If you need GitOps and you have Flux, Argo, CD, all of these different options
8:32
So that's why also I think the platform term is getting so much widespread attention
8:37
because it actually encompasses the fact that it's just not just Kubernetes
8:41
there's a host of other things there, which can also use productivity tooling as well
8:45
And all of these kinds of things, in addition to, more traditionally viewed CNCF projects
8:50
that are like Kubernetes add-ons as well. Yeah, yeah, good point. And so a couple of things spring to mind here
8:58
One is that Microsoft is, for example, not telling us exactly how they are building
9:03
the higher level services that I like to call them, the services that we are seeing now in AI and ML
9:10
All of those are built with a lot of building blocks. You have storage, you have messaging
9:15
you have all kinds of things, data ingestion, pipelines and things, and of course compute
9:21
So I'm sure that Microsoft are running a bunch of automated scalable clusters
9:26
and containerizing the execution of various algorithms that you want to run as an ML, as a data scientist, right
9:34
I'm sure they're doing that. I'm sure they're using it a lot, but they're not telling us exactly how they're doing it
9:40
But that's the thing to achieve this massive scale. Exactly, and it is true
9:44
Like JTP, the open AI, there is a KubeCon keynote a few years back as well
9:49
where they've talked about how they use Kubernetes and how they utilize these platforms
9:53
So that is an interesting point to start digging into. And there's CNCF project additions there
10:00
So ML, like Kubeflow, for example, for managing all of those things where you'd need to have a self-service way
10:06
for data scientists, for example, using these. But there's so many of them
10:09
There's over like 150 or something, you know, from CNCF that you can tap into
10:14
So there's so many things that you can select. But obviously keeping in mind always
10:18
that not just run into random projects, just because they're cool, even though I love cool projects
10:24
and I'm the one who's always like, this is cool, this is cool, this is cool. But at the same time, think about what your application
10:31
what your infrastructure, what they actually need, what is truly needed for you
10:34
because you don't also want to add complexity just because, because that's never gonna end really well either
10:40
Yeah, yeah, yeah, yeah. Very, very good point. So the other thing that I thought about
10:45
from what you said before is you said that there are some parts of running Kubernetes
10:51
that are highly advanced and you can do very good things with it
10:55
So don't get me wrong. I mean, it's great, except that you have to be able
10:58
to be on top of these things. What would you say are the biggest challenges
11:03
for a company that wants to, are considering running something like Kubernetes
11:08
Oh, the technical people are coming to the manager saying, we're gonna use Kubernetes
11:12
we're gonna run our clusters. Managers don't know what that is. So what are the biggest concerns that they should be
11:18
or the company should be focusing on now? What's gonna be the hardest
11:23
Yeah, and there's actually, I think there's a, well, it depends on the maturity level of your company
11:27
and like as far as cloud native and cloud use to begin with goes
11:30
as well as the technical knowledge within your teams. But I think for me as a kind of Kubernetes
11:36
and cloud native CNCF enthusiast, I'm very deep in Kubernetes topics constantly
11:42
but it always surprises me. And I think it shouldn't really surprise me
11:46
that how much actually people still need sometimes relatively quote unquote, basic information and law knowledge on Kubernetes
11:54
But so CNCF does a survey every year where they look into what are the trends
11:59
around these usages around Kubernetes, cloud native projects and whatnot. And what are the, to me, the most interesting part
12:06
what are the challenges that people are actually facing using these products? And for years now, there's essentially been two
12:12
I think in the really like high up, either the number one and two are like relatively there
12:18
has been knowledge and tooling, like knowledge, essentially education on how to use Kubernetes in the right way
12:24
because it is complex as you said. And then the second one, which is rising
12:28
I think every year, security. So that's another side that I think is very much becoming
12:34
as well as has become a big topic. And I think will rise in popularity
12:39
and importance as we go on. But I do wanna give a shout out. They actually, I think released one of the new state
12:46
of CNCF and Kubernetes surveys like today. I haven't had the time to check it yet
12:50
So I have no hot off the press news, but like as a tip for everyone that if you're interested
12:56
these are kind of being published all the time. Well, once a year, but there's different areas
13:00
that you can check into their spin-offs, their security and whatnot. So highly recommend checking these out
13:05
if you're looking for, I think, peer support on, are you alone with your struggles, for example
13:11
Okay, cool. So that's great. Now, the sweet spot of using Kubernetes
13:19
I'm assuming you're gonna go with something like when it's automated, when you use DevOps and things
13:23
and that's okay. You can go with your baby. Like you can go with what your company does
13:28
and so forth, right? So when you really are in the flow, when you're in the zone, how good is it
13:33
What is it that you're doing right when you're doing it the right way
13:37
Like, ooh, sweet spot. Well, I think for me, at least with technology
13:42
with any technology, both with these and as well as anything, if things work, I'm happy. Yeah
13:50
Like that's kind of the minimum requirement and also the best case scenario, which is funny to kind of say
13:57
but I think that's kind of what it is. And obviously, like we can go into making things super optimized
14:02
and I do actually talk about that as well. And I think that should be considered for, you know
14:08
sustainability and cost optimization is a topic as well. But I think if those are hardwired into your requirements
14:14
on what your platform and your Kubernetes usage should be like, so cost-effective or sustainability
14:20
then that should be included in the, it works, you know, requirement
14:24
That is the requirement. And I think no one who uses Kubernetes is no stranger
14:29
to, you know, battling some weird error codes here and there and, but trying to figure out bugs and everything
14:35
But it's, you know, as I said, if it works, it works and I'm happy then, so that's it for me, at least
14:41
That's it for you. Okay, so you don't have, I mean, I guess that's a movable goalpost, right
14:46
Once you have figured some things out, there's always something new to optimize
14:51
or change and so forth, I guess. Yeah, and I think it's because I am a big proponent
14:55
for essentially like what makes sense for your infrastructure, your company and your situation
15:00
Like people, if I, so for example, I speak a lot about KEDA, which is auto-scaling
15:05
which I think it makes sense, like just from a very like logical standpoint, auto-scaling
15:11
So like, you know, you're saving money, you're saving environment, like resources as well
15:15
You're saving time, you don't have to do things manually. It all makes sense, but a lot of times
15:20
after I speak about it, people might come up to me after the conferences and talk about, you know
15:24
oh, for these and these reasons, we couldn't use it and they look a bit sad and I'm like, well, it's fine
15:31
Like you had your reasons, it makes sense. I think it's fine
15:35
No one has to force themselves to use it if it doesn't make sense for their use case for a variety of reasons
15:40
So, and I think that you just have to kind of, you know, deal with that as well
15:45
So it is very much a movable goalpost so you can figure out what works for you
15:49
in terms of what actually means it works for me. Yeah, yeah, definitely
15:54
So I guess I don't wanna, you know, end on a note
15:59
that says that you shouldn't use Kubernetes, but you know, what are some cases where people are just
16:05
they're getting in over their head. I mean, they shouldn't really be doing this
16:10
You were saying the scale thing initially, right? Yeah. If you don't have that scale
16:14
then why are you using Kubernetes? You should be using a web app, right
16:21
Is that like the only reason to not, you know, go for this, you know
16:25
highly capable and advanced machinery? I do honestly think the scale is often the root cause
16:31
for these scenarios. Yeah. Because particularly from the level of like, you know
16:36
Kubernetes is built for scale. If you're not managing that scale, it's gonna be a bit complex for you to use
16:43
But again, if you want to learn it, if you want to use it, I'm not gonna stop you
16:48
But at the same time, you know, if like, it's the same as like, okay
16:52
if you're, you know, sweeping with the broom, like you're sweeping with the broom, it's working
16:57
Do you need to have a hundred brooms at the same time, you know, working
17:01
If you just need one broom to sweep your house, no, you don't
17:06
So, you know, yeah. So these are one of the reasons. And obviously that, because that brings all the scale
17:11
like the scale means that it brings the complexity and it means a lot of different managing issues
17:16
and everything. So then you're kind of creating more issues for yourself than you're solving necessarily
17:21
But then you have obviously levels to it. So you can start using Kubernetes relatively nowadays
17:27
quite easily with AKS, EKS, you know, one of these. But at the same time, there's a lot of people who I talk to
17:34
who use Kubernetes, but then they might say that, oh, Kubeflow itself was a bit too, like added a bit
17:39
too much complexity to our, you know, ML, AI scenario. But Kubernetes is a good backend and infrastructure
17:45
that they can use. And I think that's also like the other level, like, yes
17:49
sure you can use Kubernetes for the scale that you have, which might be the threshold that you might meet
17:54
quite easily. But then we're thinking about, you know, Kubeflow or any of these like add-ons on from, you know
18:01
Argo CD, GitOps, whatnot, or any of these sides, you know, some people need multiple different type of service meshes
18:07
Would I recommend it? Not necessarily, but if you do, then you might need to plug in Measury
18:13
which is a service mesh management plane. So then you get into these situations where to solve
18:19
for the scale or complexities that your application brings, you're essentially adding more complexity
18:24
which makes sense. But also if you don't fundamentally need these services
18:28
you're just adding complexity for the sake of complexity and will probably run into some issues down the road
18:34
Yeah, definitely. So you're an Azure gal, same as I'm an Azure guy
18:39
We have that in common. So what would you be calling out in terms of other services
18:47
that are like your favorite services to use together with your clusters and stuff like, you know
18:52
application insights and what else commonly goes with a Kubernetes deployment? Well, I think there's too many to mention
19:01
And as an Azure MVP, this is honestly the most stressful. Exactly, this is the most stressful question for me
19:07
because I feel like I want to mention everything, but then I'm like, I don't know. Yeah, but it depends on your use case
19:14
Like I am nowadays playing around a lot with Azure AI services as well
19:18
and open AI connections there and everything, mostly because that's kind of where my demos are
19:23
for example, at the moment, but there's so many. And to be frank, like I think, and Vision as well
19:30
we offer a lot of cost optimization services. So from that side, actually just, you know
19:34
basic billing help and, you know, advising services, you know, are actually really useful
19:41
that I think people overlook quite often for basic management. And to be also honest
19:47
policies are a big one that people quite often overlook, both for security and for costs as well
19:53
and for like sustainability. So I do like a lot of these that are in-built
19:57
in Azure as well. So you have to measure, same as with everything else
20:01
You need to measure your cost. You need to measure, well, I mean, to monitor if you are secure
20:07
and you need to measure how much usage you have in your cluster and so forth
20:12
All right, that makes a lot of sense. Well, I think that's it for time for this episode
20:18
And I think it was a really, really good one. I think it's a strong, you know
20:21
a walkthrough of how does this even work? What should you be thinking about
20:27
And I guess security and automation and scaling. Yeah. Like the big ones
20:35
These are big ones, yeah. And I think it's always depends on the, I keep saying it depends on the scenario
20:40
but it really does. So it's the point. It depends. It is, it is. It's the favorite answer from any consultant
20:46
Yeah, it is. But I do think you can really see, for example
20:51
the need for auto-scaling from how fast KEDA has moved. Well, it's always relative
20:57
but KEDA has gone from like a sandbox to graduated project, for example
21:01
And you can see the people are really using it, loving it and so forth
21:05
Or to be honest, there's obviously the cluster auto-scaler and all of these things as well
21:09
But these are a lot of use, like user-scale scenarios for Kubernetes
21:14
because it's built for scale. But then auto-scaling really helps with a lot of the cases
21:18
And I think, yeah, security is a big one as well. And we really do have to pay attention to it
21:22
And so the trend is to work towards getting to a place
21:26
where your Kubernetes scenario feels like a platform service scenario. That's what you want to get to
21:34
You shouldn't be thinking about that other stuff so much. Exactly. That is the trend currently
21:40
and it is gaining traction constantly. And you can see it in KubeCon programs
21:45
You can see in the way people talk about these things. There's so many good resources around that as well
21:51
and so forth. So there's a lot to dive deeper into there as well
21:55
if anyone's interested. All right. Well, thank you so much, Annie, for being on the show today
22:01
And thank you everyone who was watching the Cloud Show, and I'll see you again next week
22:06
Thank you for having me. ♪♪��

The Cloud Show with Magnus Mårtensson ft. Annie Talvasto - Ep. 19

CSharpCorner

Small Business Trends Hosts a Zoholics 2024 Roundtable #zoholics24 #zoholicsUS

The Cloud Show with Magnus Mårtensson ft. Hanna Engel - Ep. 17

The Cloud Show with Magnus Mårtensson ft. Martin Thwaites - Ep. 22

The Cloud Show with Magnus Mårtensson ft. Rik Hepworth Ep. 3

The Cloud Show with Magnus Mårtensson ft. Dr Greg Low - Ep. 11

Starting with Azure... - Azure for Sure Ep. 1

Bringing DevOps to the Database by Steve Jones || SQL Server Virtual Conference

The Cloud Show with Magnus Mårtensson ft. Logan Song - Ep. 23

The Cloud Show with Magnus Mårtensson ft. Vishwas Lele - Ep. 12

The Cloud Show with Magnus Mårtensson ft. Joseph Guadagno - Ep. 9

Copilot in Dynamics 365 CRM Sales - Microsoft Business Application Live Show Ep.18

Benchmarking in the cloud by Gianluca Sartori || SQL Server Virtual Conference

The Cloud Show with Magnus Mårtensson ft. Robert Folkesson - Ep. 20

How to Upload Files in Autodesk Construction Cloud (ACC) | Quick Start Tutorial

How to Install Ansible AWX on Kubernetes Cluster | Open Source Ansible Tower

Up next in 10

The Cloud Show with Magnus Mårtensson ft. Annie Talvasto - Ep. 19

CSharpCorner