Node-Level Optimization in Large Cluster Powered by Containerd Plugin
Talkat KubeCon + CloudNativeCon + Open Source Summit China 2023 · on Wednesday, 2023-09-27· at 14:45· 3层 305A会议室| 3F Room 305A·
summary
Describes using containerd plugins plus lazy-loading (OCI) images and P2P image accelerators to reduce pod cold-start and mitigate IO “noisy neighbor” effects in large Kubernetes clusters. Concretely, it covers a dynamic image garbage‑collection plugin and techniques such as separating read/write disk paths to manage IO utilization for multi‑tenant, latency‑sensitive workloads (e.g., Function Compute).
AI-written summary · about
related talks
36mOptimizing Kubernetes for Low-Latency and Real-Time Containers
20mFrom Cold Start To Warp Speed: Triton Kernel Caching With OCI Container Images
37mCache Me If You Can: P2P Image Sharing in Kubernetes with Spegel
01h 17mIntro + Deep Dive: containerd
35mThe Growing Need for Caching Container Images in Kubernetes
30mFaster Container Image Distribution on a Variety of Tools with Lazy Pulling
11mFast and Efficient Container Startup at the Edge via Dependency Scheduling
32mFive things you need to know about your container runtime