Aflevering 132: From CPU to GPU: The New Reality of Kubernetes 1.36

In this episode, Ronald and Jan are joined by Nigel Douglas, Head of Developer Relations at Cloudsmith, to discuss the upcoming Kubernetes 1.36 release and the broader evolution of the Kubernetes ecosystem.

Nigel shares his journey from help desk and cybersecurity roles into open source, eventually working closely with Kubernetes through projects like Calico and Falco within the CNCF ecosystem.

The conversation centers around Kubernetes 1.36, which marks a shift from foundational features toward optimization and new use cases. A major theme in this release is the growing importance of AI workloads. Kubernetes is increasingly positioned as the orchestration platform for AI, with features like Dynamic Resource Allocation (DRA) enabling better management of GPUs and specialized hardware.

Security is another dominant theme. Many of the changes in this release focus on closing gaps and improving control, such as more fine-grained authorization, better admission control during node startup, and addressing previously existing vulnerabilities.

Additionally, the episode highlights several practical improvements, including better snapshot capabilities for stateful workloads, enhanced observability features like native histograms, and improvements in workload scheduling that take hardware topology into account.

The discussion also touches on a common challenge in the Kubernetes world: upgrading. Many organizations still run older versions due to the complexity of dependencies and ecosystem changes, making transitions non-trivial.

Looking ahead, Nigel emphasizes the need for more standardization within Kubernetes to make it easier for organizations to adapt when components change or become deprecated, reinforcing the importance of a stable and predictable ecosystem.

‍