Understanding Scalability and Performance in the Kubernetes Master
summary
Benchmarks and profiling of Kubernetes control plane components (etcd, API server, scheduler) to identify which become bottlenecks beyond the documented 5k-node limit. Describes concrete tuning and master-side optimizations applied at Alibaba that sped up pod creation and other control-plane operations by over 100×, enabling smooth operation of 10k-node clusters.
AI-written summary · about
related talks
36mUnderstanding Scalability and Performance in the Kubernetes Master
34mPanel: Improving and Managing Kubernetes at Scale
48mScale and Performance Testing of Kubernetes: Answers for Specific Applications
02h 26mKubernetes Performance Tuning Workshop
34mScaling Kubernetes up to 1,000 Control Planes - Dario Tranchitella
32mScaling Kubernetes Application to 1M Users
34mA story of managing kubernetes cluster with 15k nodes & various workloads
50mScale Kubernetes to Manage 60K Nodes Through Architectural Extension