What if we stopped defaulting to one cluster per region and started thinking of Kubernetes as one giant, global substrate?

This talk explores how we challenged conventional cluster architectures by running Kubernetes worker nodes across geographies from a single control plane. In edge and resource-constrained regions, deploying redundant control planes is both expensive and fragile. Instead, we designed a global worker-plane architecture, gaining a unified control surface, simplified observability, and efficient resource sharing.

We’ll share architectural patterns, lessons learned around latency, resiliency, and networking, and how this approach reduces operational sprawl. Attendees will leave with practical considerations and a new lens on cluster design.