Understanding Scalability and Performance in the Kubernetes Master

Explore the scalability and performance challenges of the Kubernetes master in this informative conference talk by Xingyu Chen and Fansong Zeng from Alibaba. Delve into the current scale limit of Kubernetes at 5,000 nodes and investigate the performance bottlenecks preventing management of larger clusters. Learn about Alibaba's experiences with slowing pod creation in growing clusters and discover their approach to benchmark testing and profiling. Gain insights into the tweaks and tunings applied to the master that resulted in a remarkable 100x performance improvement, enabling smooth operation of a 10,000-node Kubernetes cluster. Understand the key factors affecting large-scale Kubernetes cluster management and how to identify and address potential limitations in etcd, apiserver, or scheduler components.