A Huge Cluster or Multi-Clusters? Identifying the Bottleneck
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 15:00· Level 1 | Hall Entrance S10 | Room A·
summary
Compares concrete limits and fixes: tested guidance for 5,000+ node clusters, Google’s 65,000-node demo, ByteDance’s Kubebrain at 20k, and when to split into multiclusters. Covers actionable knobs: API server rate limits, APF, read caches, kube-proxy NFtables, and etcd tuning plus replacements like Kubebrain/kine; shows vcluster’s virtual control plane demo and quark for scale testing, and surveys multicluster tools (Karmada, Clusternet) and networking pain points like Submariner.
AI-written summary · about
chapters (15)
- 0:00Welcome
- 1:52Kubernetes cluster sizes survey
- 2:57Cluster scaling benchmarks
- 5:49Large cluster pros and cons
- 8:15Large cluster challenges
- 12:12Kubernetes scaling updates
- 14:47Reducing etcd pressure
- 17:00Alternative cluster solutions
- 19:30Key takeaways
- 20:39Namespace multi-tenancy limits
- 23:25Multi-tenancy solutions spectrum
- 26:18Control plane as a service
- 28:26vCluster introduction
- 30:35vCluster demo
- 33:04Closing remarks
related talks
34mPanel: Improving and Managing Kubernetes at Scale
42m2000 Nodes and Beyond: How We Scaled Kubernetes to 60,000-Container Clusters
30mScaling Kubernetes to Thousands of Nodes Across Multiple Clusters, Calmly
37mOne Cluster to Rule Them All - Multi-Tenancy for 1k Users
31mThe Myth of the Monocluster
03mFrom 5 to 1,300+ Clusters: Declarative Scaling for Private Kubernetes
33mKubernetes Multi-tenancy: Principles and Practices for Large Scale Internal Deve...
11mKeynote: Multicluster Strategies to “Kubernetify” Legacy Apps