Sponsored Keynote: Open Source Innovation – Success Through Failure - Thomas Di Giacomo, SUSE

Thank you for joining, and welcome to this short talk about open source innovation in the context of CNCF. Obviously, this is the first KubeCon for us since SUSE and Rancher came together — two strong open source DNAs merging into one. And as one, we are more committed to and focused on open source and interoperable innovation than ever.

As Thomas Edison once said, 'I didn't fail. I just found 2,000 ways how to not make a light bulb. I only needed to find one way to make it work.' The same mindset holds true for open source innovation, where the vast majority of projects fail — at least as measured by user adoption — but that is in fact the secret of open source innovation success through rapid iterative failure and learnings.

Hopefully today I'd like to share how we massively and iteratively experiment with no rules around cloud native open source technologies to then contribute and deliver rapid innovation to the CNCF community and to our enterprise customers at any time.

We are working on many exciting open source projects. Only a few succeed and make their way back to CNCF, like Longhorn and K3s, but all of them contribute to advancing open source cloud native innovation.

SUSE and Rancher engineers are constantly launching and starting new open source ideas and projects based on CNCF technologies and additional cloud native needs that we are seeing ourselves or that are coming from users, companies, or developers in general. This approach is in our DNA. We encourage it, and we are intentionally structured accordingly.

It is very grassroots, community-inspired, and happens both on-mass and in very different locations of the cloud native landscape. There are no rules except these: the projects are of course open source, they are accessible to everyone, and by design they are independent of specific architectural choices. This is to ensure total flexibility, choice, and freedom for how the projects can be used, assembled, and developed.

Our current exploration projects are broad and diverse, ranging from infrastructure and policy management to applications. For instance, the CubeWarden project is facilitating policy creation and management through a WebAssembly implementation of the Kubernetes admission control.

Our project Harvester, which is our vision for the next-generation hyper-converged infrastructure, Harvester takes data center metal and produces a Kubernetes API from where one can manage it and also deploy containerized or virtualized applications.

Another example is Hipper, that experiments with enhanced dependency management concepts in Helm. I could go on for hours, and actually by the time we discussed this, many of those projects might have changed. But the point is that there are so many directions to explore that it needs to happen widely and almost chaotically, without any restriction other than being open.

As you can imagine, not all of these will succeed — at least in the sense of not ending up being a relevant component for the cloud native ecosystem. In the end, a small number of them will succeed and then join the CNCF projects for further development. That's exactly the path K3s and Longhorn have taken — initiated outside of CNCF but joining it as they have matured, gained relevance, and been adopted.

These projects have expanded nicely since joining CNCF a few months ago. Together with the upstream containerd and SELinux communities, SELinux support to containerd was introduced and incorporated into K3s — another example of K3s evolution thanks to the CNCF community — is the support for embedded etcd as a data store option, so that K3s clusters can easily scale out from single nodes to multi-node, highly available clusters.

On the Longhorn side, ARM64 support has been added since Longhorn joined CNCF. The community saw use cases where a Kubernetes-native distributed block storage like Longhorn would be relevant for edge and industrial computing, where ARM is obviously very present.

Self-healing capabilities were introduced too, enabling Longhorn to automatically take action with Kubernetes to relocate pods and their data from a failing node to a healthy one.

And as a last example, there is the read-write-many support. The community has extended the block storage foundation of Longhorn, adding a network file system on top of it to allow multiple nodes to read and write from the same volume.

Thank you everyone for the interest and the contributions to K3s and Longhorn. If you'd like to know more about the exploration projects too, there are sessions here at KubeCon. You're also very welcome to have a look at them on our GitHub repos and to reach out to us directly. We are more committed than ever to open source innovation around cloud native, focusing on choice, flexibility, and freedom to put the exact cloud native puzzle together that someone would need. Thank you very much and have a great KubeCon!