Nydus + JuiceFS: Reducing Container Startup Time for AI Inference from 116s to 1.4s

In large-scale AI inference services, when a cluster scales out, newly added inference instances must...