Loft Labs introduces vCluster Cloud: SaaS for multi-tenancy

Bart: Can you tell me who you are, what your role is, and where you work, specifically at Loft Labs?

Lukas: Hi, I'm Lukas. I'm the CEO and one of the co-founders of Loft Labs. We make some cool open-source projects, including vCluster and DevPod.

Bart: What news do you want to share with us today?

Lukas: We are launching vCluster Cloud, which is the first time we are launching a SaaS product. This means a vCluster platform to manage a lot of virtual Kubernetes clusters. You can now have that platform running as a SaaS rather than having to self-host it. It's a great way to explore the platform and check out the features that we're launching there without having the hosting responsibility. It's in beta right now, but we're working really hard to make it a stable version pretty soon.

Bart: What problem does vCluster Cloud solve?

Lukas: vCluster and vCluster Cloud solve the multi-tenancy problem in Kubernetes. If you have hundreds of clusters or even 50 Kubernetes clusters, there's a lot of redundancy. You have 50 Istio, cert-manager, OPA, Prometheus, and everything is duplicated in all of these clusters. It's a huge waste of resources, especially in pre-production, since most of these components run idle all the time. We are advocating for creating bigger clusters and having shared components. The only problem with that is how to split it up and how to give people access to that shared cluster. Giving them access to a namespace may be really limiting for them, and there are a lot of noisy neighbor problems in that case. vCluster is a technology that allows you to launch a Kubernetes control plane inside a pod of a namespace, and then you can give people access to that virtual API server. They feel like they have a full cluster, similar to a VM, where they are rooted inside and can do anything, but they are actually running on a shared host. That's the beauty of it.

Bart: Could you give me a little bit of context, the before and after of this announcement?

Lukas: vCluster Cloud makes it much easier to export a commercial product than ever before. The open source version is easy to spin up on your local machine. However, if you want to connect your cloud provider resources, spin up the platform, and expose it via an ingress to open up the UI, vCluster Cloud is a great option. You can head to vCluster Cloud, sign up, and get the vCluster platform running in under a minute.

Bart: Are vCluster and vCluster Cloud open source and part of the CNCF landscape?

Lukas: vCluster Cloud is a commercial product, but vCluster itself is open source. It's Apache 2 license and we want to keep it that way. We are very honest about our company strategy. We have to build a sustainable company in order to fund our open source efforts. We do not plan to relicense vCluster. We do not want it to be like a bait and switch kind of model. We want to make sure vCluster is a solid and amazing open source project. And then we have an amazing commercial product on top of that, vCluster Cloud, as a part of Loft Labs. That's what we're all about.

Bart: What's your business model?

Lukas: Our business model is essentially layered in two phases. The core part is the open-source vCluster, which is what you see on GitHub, including all the core features. We also have pro features that are helpful when your vClusters are under a lot of load, such as specific security features and compliance things that enable running vCluster at a much more resilient and higher scale. These pro features can be unlocked with a license key. The second level consists of platform features, which help orchestrate a fleet of virtual clusters. If you have a large number of virtual clusters, questions arise about standardization, ensuring SSO and audit logging across all virtual clusters, and optimization. For example, we have a feature called sleep mode that puts idle virtual clusters to sleep when they are not being used. This is essentially what our commercial model looks like on top of the open source.

Bart: And who are your main competitors?

Lukas: I would say our main competitor is the current state of the art. A lot of people still spin up 10 Kubernetes clusters every month or so in an organization. We suggest reconsidering this approach. Instead of spinning up dedicated clusters, consider using vCluster. You do not have to migrate your 200 existing clusters into vCluster immediately, but for every new cluster you spin up, think about whether it needs a dedicated set of nodes and a dedicated Kubernetes cluster, or if it can run in a multi-tenant cluster and you can launch a vCluster instead. The current state of the art is really our main competitor.

Bart: What differentiates vCluster from some of the other options that are out there?

Lukas: I think vCluster really aims to be easy and easily adaptable. To spin up a vCluster, you don't need cluster-wide permissions. Sometimes companies come to us and say they have 400 vClusters already running, and they just found out. This is because you don't need special admin permissions in a cluster. All you need is the permission to create a stateful set and a service, and then you can launch a virtual cluster and elevate your privileges without actually elevating your privileges in that cluster. The simple way of deploying a virtual cluster and getting started with it is really the power of it. You can adopt this in about five minutes and see the results. I think a lot of other projects that aim to do multi-tenancy in the Kubernetes space are much more complicated and require a lot more permissions in your cluster. We try to have as minimal a footprint as possible to start with.

Bart: What should we expect next from vCluster and Loft Labs?

Lukas: We just shipped the KubeVirt integration, which is really exciting because it brings virtual machines to Kubernetes. We are going to invest more in bringing dedicated nodes into Kubernetes clusters, like vCluster. Traditional Kubernetes clusters have all dedicated nodes, whereas a vCluster, which we launched in 2021, has all shared nodes. Maybe there's a middle ground where certain nodes can be dedicated and others shared in a shared node pool. Depending on the workload and whether it needs extra access to a dedicated node, we can schedule them differently. This is a very exciting area that we are currently researching.