Jamming with CRI-O & crun: Facilitating the Convergence of AI, WASM, and Kubernetes
summary
CRI-O + crun run native WASM workloads in Kubernetes by detecting OCI images with platform wasi/wasm32 and dispatching them to crun with WASM plugin; Kubernetes mounts an OCI image as a volume (alpha feature Gate imageVolume) so a 4 GB model artifact is pulled once per node and shared across containers. Demo builds a scratch container that copies a llama-edge .wasm chatbot, injects host WASM-Edge shared libraries as plugins, and runs OK on CPU though latency is high without GPU/WebGPU.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:32Overview
- 7:40Work done
- 13:14Software needed
- 18:59Running CRIO
- 22:09Future Work
- 23:42References Questions
related talks
38mCRI-O's WASM Adventure: Challenges, Strategies, and What Lies Ahead
34mManaging WebAssembly Applications Using Docker and k8s Tooling
38mKubernetes on Edge: Bringing Your Code to Constraine...
25mManaging WebAssembly apps with Kubernetes
28mWasm + Kubernetes: Beyond Containers
33mKubernetes on the edge: running low-latency workloads with WebAssembly
18mCloud Native & WebAssembly: Better Together
32mWebAssembly Based AI as a Service on the Edge with Kubernetes