Misadventures in Large Scale Cluster Performance - Debunking Kubernetes Best Practices

Explore a technical conference talk that challenges conventional wisdom about Kubernetes performance optimization through real-world findings and extensive research. Delve into detailed investigations spanning two years, including analysis of Linux kernel performance, control plane configurations, and custom monitoring implementations. Learn from AWS and Lacework experts as they present surprising discoveries that contradict popular Kubernetes best practices, backed by evidence from their hands-on experimentation and monitoring. Gain insights from their deep dive into cluster performance optimization, building upon their highly-rated previous presentation on multi-tenant deployments with 73,000 pods. Understand which commonly accepted practices may actually be limiting your cluster's performance and discover evidence-based approaches to improve large-scale Kubernetes deployments.