Running Kubernetes at Scale – Sebastian Scheele, Kubermatic

This is Textron TV. Hey everyone, welcome back to Tech strong TV. We've got to actually have a new company and and First Time guest here on texture on TV to introduce you to want to introduce you right now to Sebastian Scheele.

Sebastian is the co-founder. I believe CEO. Yeah, that's right direct about you.

Yeah, okay and and the name of his company. I want to make sure I get it right here is coopermatic. And from the name, I you could probably guess it has something to do with kubernetes, but Sebastian why don't first of all welcome to text you on TV.

Thanks for being on. Why don't you if you don't mind share a little bit of your background with the audience? Yeah, first of all, thanks for the invitation.

So giving you a big of background about me and about the companies. So yeah, I'm the Bastian one of the two Founders and CEO of chromatic. I have a technology background originally studied computer science work.

Then a lot of years in the Enterprise industry latest in the product management for sap in the database space. Yeah, and when quite early into the docker and kubernetes space like I think I started. Seven and a half years or eight years ago was darker.

5 or something like really on early on on this Cloud native train and six and a half years ago. We decided when Google announced kubernetes say that's an interesting tool which is potentially not only for the Google Cloud the interesting for other Cloud providers, but also for environments, let's do something with it. And so then we started a company and we saw the opportunity.

How can we help customers to manage communities last us anywhere like from the cloud to their bigger data centers, and now more also expectually on edge Computing so having like a unified platform which helps you to manage Q&A this class before we jump more into co-metic potentially. Let's finish a bit about me. So beside here CEO for cool magic.

I'm also in the governing board from the Linux networking foundation and the cncf we need to help and drive the perspective of smaller companies in this organization and providing that feedback also to the foundation what's smaller companies? Are doing and where they want or what their opinion about where the whole ecosystem is going to? Absolutely interesting stuff great stuff.

I guess I should ask you that's just because I'm curious. Will you be a cute continent answer Amsterdam in April? Yes, of course.

I was a coupon Detroit. What was it work? Yeah, we're there too.

And next time you next time that's what I'm saying. We're gonna be in Amsterdam doing videos. So we'll do a part two of this.

In April an Amsterdam, but it's interesting. So but let's talk about the the company's offering if you will. right What what do you guys you know offering to customers and who are the kinds of customers?

Yeah, so what we do or potentially like when we started we saw the challenge how to manage kubernetes class stars at scale at that time. There was only gke available where press a button and get a kubernetes class that and we were asking ourselves. Hmm.

We want to have something similar but on any cloud provider and also on on-prem environment so technically what we build more than a platform which helps you to manage tens hundreds or even thousands of kidneys across all the different platforms. We started initially especially with managed service providers. So who who are running or providing communities as a service to their customers because one of the strong capabilities we build in from the first day on was multi tendency capability so that different customers or different users can't see the workload of the other.

And so we started with them because they had from the first day on this requirement. I cannot share kubernetes class. I was multiple customers and so they needed to manage a lot of kubernetes clusters.

And and a lot of Enterprises at that time was still at the stage. No, I'm running one or two big kubernetes cluster do multi tendency inside of the class that was different namespaces. But we we saw that the problem like if you do this the problem what you solve on the application Level you push down into the kubernetes infrastructure level because when do you want to upgrade your kubernetes cluster?

Because there's every time someone who don't have time to upgrade the application or who is currently critical workload. And so you have then a lot of flexibility in the application, but still back in the infrastructure on upgrading twice a year Christmas and Eastern or something like this you underlying infrastructure and so over time more Enterprises or working with more Enterprises because they also getting more clusters the management of this class does get more complicated because across all this different environments. They see really that they want to centralize this having a container as a service or it's not native as a service team, but then they also need to provide a platform and like that's exactly where we fit into place.

So if people start working at scale They want to build their own as a service offering also inside of the organization so that others can. Be an API consume the services they providing and our starting point is kubernetes there so that they easily can start with creating kubernetes cluster and others can operate this for them. I mean it's almost in those cases.

It's it's really having a kubernetes as a service serving an internal. Customer base, right? You're not selling it to the public or the customers.

You're out there there. It's internal. Now when you say it's kubernetes is a service.

It's not just the kubernetes. It's the whole it's the whole platform right the whole stack if you will. Yeah, it's not only like okay, you're getting in kubernetes clip class.

It's a whole platform like we take care about how to provisional kubernetes cluster. But then also how to do heavily day two operation how to upgrade kubernetes cluster how to automatically for example when you have when notes go down how to recreate them automatically, but then things on top like for example how to ensure that you can roll out policies for example policies on the Clusters, especially like an Enterprise this is coming up more and more that you say okay for production classes. I have a certain set of policies which needs to be applied that I can serve them as a production class that or then we have capabilities of And application catalog so that your platform team can provide a decent set of default applications.

For example, an engine X or some specific combinations of application, which you're developers can easily consume and say, okay, please install this additional components on top of the cluster and one thing what we doing also want more is then providing holistic monitoring a platform so that you're not that not you don't need to install on every class that dedicated from ecos and grafana so that you can centralize this and your container or Cloud native team can also provide them out of the box stack so that connect all the classes to a central monitoring and logging infrastructure based on Prometheus or grafana and Loki so that you can centralize all the logging and monitoring information. Excellent, you know? Would it be fair to say like what you're offering is almost a distro?

a full-on distro or not And yes on the one side it's it's a distro, but it's it's even more it's like really like then the whole platform. Um how to manage kubernetes clusters at scale so that it's really like also around the management environment. And I mean, we have two tools one is like our chromatic humanitas platform in short like.

Kkp which is really like this how to manage kubernetes at scale. And when we started building this we were thinking okay. What's the best way to manage kubernetes cluster and we came up with the idea.

Let's use kubernetes to manage kubernetes, which is nowadays called kubernetes operator, but when we started there was nothing there. So we like a whole organization. What we doing is from the first their own building kubernetes operators, but then we hadn't changing egg problem because we need Cuban eaters to get KP running.

So we have another tool there which is called q1, which is there to spin up individual classes and expressly for customers or for developers who only need one or two or three classes. It's a really easy starting point getting more or less the same capability as the big KP, but it was only for an individual class that and which must much less complexity because it's only their phone individual class that not like okay. I need to have the whole big orchestration engine to manage hundreds or thousands of classes at scale.

Got it. excuse me, so I think we understand a lot about the company now. Let's say Sebastian, let's let's turn our attention to You know kubernetes.

It's scary. Look. Kubernetes, no one ever raises their hand and says kubernetes is easy to use or very easy to you know, maintain or whatever kubernetes at scale is a whole different animal right?

It's certainly not for the faint of heart. What are some of the biggest problems that you see from your customers when they try to do kubernetes at scale one of the challenges? Yeah.

I think as I said like Cuban eaters itself to manage it's already hot. But if you need to manage this at scale, I think the only way how are you really can do this is with heavy automation. So if you haven't automated your whole stick and not only for provisioning your class.

I like really to do full life cycle management of your classes, like upgrading scaling scaling up scaling down deletion of class that You you run into problems and that what we see in a lot of cases where people started building things on their own it's working until a certain size. But then if you want to meet or if you need to manage more clusters, you need more Automation and that's where we come into the game like because our platform was designed and built from the first day on like how can we manage kubernetes cluster fully automated because we leveraging heavily kubernetes there. It's easy to operate a lot of kubernetes class like we did.

There's one of our customers we measured how much time they spend to operate all of the kubernetes Clusters and they spend less than five hours per month on really manual work to do it because everything is automated when something goes down our tooling and The Operators are kicking in and recreating all the stuff as it was a desire before. excellent five hours a month sounds Fantastic, I mean when you you know when you realize what you're dealing with there. Yeah, absolutely.

I mean typically we see there that even if you are already quite optimized in a lot of cases, you need one engineer potent cluster. And so if you get starting scaling more you need more more engineers in your team to operate all of this clusters and like five hours for really a bunch of clusters is nothing and nothing at all. Yeah, so you can really free up your resources.

And I mean one of the challenge everyone from us has currently is like really filing this High skilled people who knows how to operate and how to build a platform and how to manage a platform in the community space. And so instead of only using them to build this platform and able your team that they really can move application on top because that's the other challenge what we're seeing even if customers have hundreds of of us, they typically have a handful of Engineers who really deeply understand kubernetes and take this really resources to enable all of the rest of your developers so that they build application which are running smoothly on kubernetes. Got it.

You know what Sebastian occurs to me look some of our audiences are very technical audience and they they know Cloud native, but there might be some people out here who maybe really don't understand kubernetes or they're not sure when we talk about a kubernetes cluster. Right kubernetes orchestrates containers. So almost by definition when you're running an instance of kubernetes.

You're running multiple containers right a cluster. Right. Why don't you could do it better than me.

I'm sure why don't you explain what we mean by a kubernetes cluster and the multiple clusters and how that kind of You know maps to a modern not just an application because it's multiple applications but to an organization. Yeah, so I mean when we're looking into the past typically how we were running software. We were deploying it onto a dedicated individual server, even if the software was able to scale then we deployed the potentially to two or three or four servers but in the majority of cases were even having hard coded like okay, which server it goes and here first of all Container comes in containers really like how to package the software similar as what happens in the shipping industry make it an Eric.

You have one standard container. It doesn't matter what's in there. You can run it everywhere.

You can ship it everywhere and even Haber in the world can pick it up. So similar containers, you can put in Java application. You could put in go PHP no Jazz application.

And then if you have to container how to distribute this across multiple services Is really there building out of your multiple service a cluster so that they from developer perspective or from an operation perspective. You talk to One API, which is kubernetes and then kubernetes takes care automatically where to put in the individual container and what happens for example, if one server goes down that the containers were on that server are recreated on another server we have to recapacity is available and that's like in a national what company this is doing, but then the challenge comes to okay. What is if I need to have different teams what different?

Application with a different requirements for my underlying infrastructure and they are like really the Smoothie cluster approach is coming into the picture. So for example for some applications, I need to have servers with high memory or high CPU so I can put them into a closer for others for example machine learning workloads. I need gpus.

So I have a class that we're gpus are available. And so then I get much more flexibility and also can better decide when it's time to upgrade what version of kubernetes and the container one time. I'm running on this class and we're actually Pro what?

excellent good So that's your we're overtime actually ready for people who want to get more information on googlematic. Where can they go? com, there's a can find all the information or if they want to understand more.

Feel free to look at LinkedIn Sebastian Scheele. You finally there ping me and then I can help you and guide you to the right information or bring in contact with one of our folks who can provide you more information. Fantastic Sebastian.

Thank you so much for joining us today. We have a date in April in Amsterdam. I expect to see you there.

We'll continue the conversation. Absolutely. I will count you on that.

All right? Cooper manic here on Tech shark TV. We're gonna take a break now we'll be right back with another guest.