Intro + Deep Dive: Kubernetes SIG Scalability

Conference: KubeCon + CloudNativeCon Europe 2023

2023-04-19

Authors: Wojciech Tyczyński

Summary

Tips for dividing workloads among multiple clusters in Kubernetes

Networking is the most stressing for the control plane and where the biggest number of issues are seen
Understanding the size of churn forward or observed services is a significant factor in workload division
The current scalability limit of 5000 nodes is not a hard limit and there are no plans to push it further in open source
External factors like third-party controllers and ecosystem improvements need to be addressed
Using the watch protocol for getting large collections of data can help with memory consumption and system throughput
Graceful shutdowns can prevent the control plane from being blown out by hundreds of thousands of watches
Optimizations should be balanced with complexity versus return on investment trade-off

In one case, the control plane was blown out when hundreds of thousands of watches were re-established at the same time after an upgrade. This was improved in the 1.27 release.

Abstract

This session will focus on the different efforts that SIG Scalability is involved in: defining what scalability means for Kubernetes, driving performance improvements, maintaining infrastructure for scalability testing, guarding Kubernetes against performance regressions. Cooperation with other SIGs is an important aspect of the presentation as many improvements driven from the SIG are in fact owned by other SIGs. Time for Q&A will be reserved at the end of the session to understand how the SIG can better engage with the community as well as to allow the audience to provide the input about the roadmap.

Materials:

Tags: