Virtualizing Kubernetes Clusters to Boost Productivity

In this episode of View With Vizard, fresh from picking up $4.6 million in financing, Loft CEO Lukas Gentele talks with Mike Vizard about how virtualizing Kubernetes clusters increases IT productivity. The video is below followed by a complete transcript of the conversation.

Announcer: This is Digital Anarchist.

Michael Vizard: Hey guys, thanks for the throw in here with Lukas Gentele who is the CEO for Loft. They just picked up $4.6 million in additional financing to drive virtualization on Kubernetes clusters and I’m going to let him explain how all that works starting right now. Lukas, walk us through this a little bit.

Lukas Gentele: Hi Mike. Thank you so much for inviting me to talk to you about this topic. Yeah, essentially what we’re doing with Loft Labs is virtualizing Kubernetes and enabling companies to handle these virtual Kubernetes clusters in a self-service fashion so that their engineers can essentially get access to Kubernetes whenever they need it, to validate new features, to run integration tests themselves, to integrate Kubernetes into their CICD workflows.

We really want to take Kubernetes adoption to the next level. When companies tell you today they adopted Kubernetes, they are talking about, you know, maybe 20, maybe 30 people having access to Kubernetes. We’re thinking about every engineer in your engineering organization being able to spin up in self-service fashion a Kubernetes cluster, a virtual Kubernetes cluster whenever they need it for their work.

Michael Vizard: And is this going to be like one big giant cluster that I’m slicing up in the different segments ’cause a lot of people seem to have a path where they’re rolling out a lot of smaller clusters as well and so what’s going to be the mix or when do I use that in what scenario?

Lukas Gentele: Yeah, I mean we’re definitely allowing you to do multitenancy within Kubernetes, virtual clusters are a key part for a lot of companies to actually solve a lot of multitenancy challenges because virtual cluster essentially kind of in the middle between second clusters and main spaces as slices will then share a cluster. Right? Virtual clusters combine the benefits of both worlds.

On the one side, you get the great isolation that you would have with separate clusters, but on the other side, they’re also very, very cheap just as if you were slicing up your clusters. And of course it’s really unrealistic to assume that there would be only like one underlying check _____ cluster. You’re still going to need multiple Kubernetes clusters. But you need a lot fewer Kubernetes clusters, so essentially instead of spinning up a Kubernetes cluster per team, you could spin up a Kubernetes cluster for business unit or for a geographical region, for example.

That actually makes a lot of sense because you want to reduce latency, you know, someone in India may not want to work with Kubernetes clusters in North America just because the round trip through the internet and all across the world. Yeah, there will be far fewer Kubernetes clusters and then on top of these clusters, you can spin up obviously isolated name spaces, but also virtual clusters if you need more than just a single name _____.

Michael Vizard: Will I be able to provision those beforehand essentially or how long does it take to set up a virtual cluster and do I have to kind of figure that out beforehand or can I do it more on a non-demand kind of fashion?

Lukas Gentele: You know, it’s really, you know, up to the organization, but we designed Loft and virtual clusters in a way that they can be spun up very, very quickly in a self-serve session. We’re heavily betting on K3S as a very lightweight Kubernetes distribution. Our default is K3S rather than the actual Kubernetes API server. And we are using SK Light as a default storage backend to make up essentially reduce the time to spin out these virtual clusters to a minimum.

So typically it takes about 20 seconds to spin up a virtual cluster, which is much, much faster than, you know, most call providers, which take about 20 to 30 minutes or so to spin up a real Kubernetes cluster. But of course you can also if you want your virtual clusters, you know, for example, you’re testing things in your staging environment or you’re even thinking about putting virtual clusters in your production system down the road, then of course, you can also use a full-fledged SCD even to back up a virtual cluster rather than a light weight SK Light database. But again, it’s up to the use case.

Michael Vizard: Do you think folk are a little intimidating by Kubernetes clusters? It seems like we don’t see as many in production environments as I might have thought by now, but what’s your sense or where are we on that maturity curve?

Lukas Gentele: Yeah. I think the confidence is definitely increasing. I think there has been a lot of experimentation regarding Kubernetes. You know, just because – just looking at _____ landscape, right, there’s like thousands of projects, so obviously you know companies need to evaluate a lot of different technologies and figure out what is the right kind of mix that they need.

But I do feel that we’re heading in the direction where companies gain confidence to build out that spec for themselves and then to essentially, you know, put Kubernetes in more and more areas. They may just use it to start out, you know, spin up preview environments or run parts of the integration tests, but at least the customers that we have are already starting to use Kubernetes in production and you know, there’s a clear trend to move development to Kubernetes as well, which is obviously a huge chunk of the engineering organizations I’m dealing with.

Michael Vizard: Now, is your platform open source and if so, where does it fit in the landscape of the Kubernetes projects you were talking about? Are you looking at the CNCF as something you might work with or where do we go from here?

Lukas Gentele: Yeah. We are on CNCF. And are a Linux Foundation Member. At _____, I’m actually giving a talk about virtual clusters in _____ North America in LA I think in two weeks is when it starts. So we’re definitely very, very active in the CNCF ecosystem. Loft itself is a commercial product, so it’s not open source, but it is built in open source technologies. We have four open source projects out there. We have _____ cluster, obviously, which is our certified Kubernetes distribution for spinning up our _____ Kubernetes clusters.

We have Kiosk, which is a multitenancy extension from Kubernetes. We have Dev Space, which is our oldest project. It’s a development tool for Kubernetes that runs client only and really helps you streamline your dev workflows across the organization. And then we have JS Policy, which is one of our newest projects that’s essentially a policy engine that allows you to define mission control policies in Kubernetes by using JavaScript or the type script.

And our commercial product, Loft, you know, ties all of these open source projects together in one comprehensive platform, but of course you can decide which features to enable. Let’s say you want to use Open Policy Agent rather than JS Policy. You can just disable JS Policy and still use the commercial product but using _____ Cluster or Kiosk in that space. For that component, JS Policy, would disable in that case. We really built the platform to be mix and match and as customizable as possible.

Michael Vizard: Where do we go from here, I mean, now that we’ve virtualized the cluster, which in itself is something of its own attraction, what’s left to do?

Lukas Gentele: That’s a good question. I think there are a lot of challenges regarding developer experience and actually managing applications that run on top of Kubernetes and enabling developers to do more themselves ’cause right now, I think actually it’s very interesting if you are looking at the evolution of Docker versus Kubernetes. Docker was very quickly on every engineer’s machine. Right? I’m pretty sure most engineers today have Docker desktop installed or you know, some kind of Docker _____, they spin up containers and work with containers. But Kubernetes was pretty heavily focused on the operators rather than on the developers. And I think more and more developers are going to make that transition to Kubernetes because right now, it feels like we’re talking a lot about DevOps as a paradigm in bringing development and operations closer together and I feel like Kubernetes, with its heavy focus on operations and with its vast complexity, has scared off the dev part a little bit and focused too much on the ops part and I there are a lot of challenges regarding developer experience and kind of pulling that dev closer to operations again.

Michael Vizard: How automated do you think that can all get? I mean, we have self-service, so will I just kind of as a developer pop in one day and say, “Here’s what I need,” and then the whole thing will know what to configure and set up? I mean, are we that close already or will there be different profiles and different optimizations based on what type of developer I might be? Take a guess.

Lukas Gentele: Yeah, I mean a lot of people are talking about, you know, Kubernetes becoming more of the – going more in the background and almost talking about we’re heading back towards classical platform as a service, I guess and Kubernetes just has the building blocks for building these platform as a service systems.

But I think there’s a key difference. I do think a lot of things will be out of the box and, you know, less visible for the engineer at first, but I think the big difference is, when you look at all the tooling, the Google, Microsoft and, you know, companies like us or Rancho is currently building, everybody is building transparent abstractions and not the old like platform as a service where you can’t get through if you actually need to dig deeper.

And I think that’s the beauty of it. If I need to configure my readiness and lightness probes for my pods manually, or I need to look into that container and stream the logs, people would still hopefully get that CTLX _____ and that capability to twist around _____ right. But I assume most of that will be automated to just make the workflow much more efficient, but it should be automated in a transparent way rather than intransparent.

Michael Vizard: Good. Hey Lukas, thanks for being on the show and sharing your knowledge and insights. It looks like my abstractions will soon have abstractions and we’ll go from there. I want you to stay safe and we’ll see you next time.

Lukas Gentele: Thank you, Mike. Thanks for having me.

Michael Vizard: All right guys, back to you in the studio.

[End of Audio]