Explore the intricacies of measuring Kubernetes network performance in this 29-minute conference talk by Kornilios Kourtis from Isovalent. Gain valuable insights into the challenges of benchmarking K8s networking and learn practical strategies for accurate performance evaluation. Discover how to set up and execute benchmarks, avoid common pitfalls, and interpret results effectively. Delve into various workloads and metrics that address different performance questions, and understand the complexities of both software and hardware stacks. Learn about the STREAM and ECHO-REPLY methodologies, scaling workloads, and essential metrics to consider. Examine system configuration best practices, experiment design, and gain insights from a real-world Cilium case study. Acquire knowledge about useful tooling and resources to enhance your K8s network performance measurement skills.
Overview
Syllabus
Who am I? What do I do?
Why measure K8s net performance?
The stack
Evaluation process
Evaluation goal
STREAM
ECHO-REPLY: N reqs in-flight
Scaling workloads
Metrics
System configuration
Experiment
Cilium war story
Tooling
Resources
Conclusion
Taught by
CNCF [Cloud Native Computing Foundation]
