Three Approaches to Speed up Image Distribution in Cloud Native Era
summary
Covers three concrete methods Alibaba used to speed image distribution at web scale: P2P distribution with CNCF Dragonfly to reduce registry bandwidth and distribution time; using the containerd remote snapshotter to let the engine read image contents over the network without full local pulls; and a hybrid approach that dynamically loads image data from remote to local storage on demand based on read/write requests. Also compares trade-offs (e.g., network stability) and gives guidance on choosing among these approaches.
AI-written summary · about
related talks
34mThree Approaches to Speed up Image Distribution in Cloud Native Era
27mDragonfly: Make Image Distribution Efficiently and Safely in Cloud Native
29mPull Once, Scale Everywhere: Fixing Image Pull Bottlenecks With Dragonfly
34mFast Image Pulls Using IPFS And Opportunistic Caching
30mSpeeding Up Analysis Pipelines with Remote Container Images
32mCloud Native Progressive Delivery
22mKeynote: Cloud Native at Alibaba Scale
40mNode-Level Optimization in Large Cluster Powered by Containerd Plugin