2024-05 - Nephio: Scaling the Kubernetes API to support large-scale networks

60m, @Ciaran Johnston Gergely Nagy

Nephio leverages Kubernetes as the automation control plane at each layer of the stack in order to simplify the overall automation and enable declarative management with active reconciliation for the entire stack. Managed resources are modelled in KRM and many are custom resource definitions that are created in the Kubernetes API server in the management cluster to enable active reconciliation by operators.

There are some concerns over the scalability of this solution for certain use cases. To bottom these concerns out, it is worthwhile looking at the potential numbers of CRs that will be created and defining some test cases to verify the upper bounds of those limits before we start to reach them in the wild.

In this session I will aim to propose some numbers - theoretical and possibly practical - and trigger a discussion on the future way forward.