[{"data":1,"prerenderedAt":17},["ShallowReactive",2],{"blog-tag-ja-device-plugin":3},[4],{"path":5,"title":6,"description":7,"date":8,"tags":9},"\u002Fblog\u002Fja\u002Fk3s-edge-ai-inference-fractional-gpu-time-slicing","K3sでエッジAI推論基盤を組んだら、1台のGPUに2つのモデルを乗せられなかった。fractional GPUという壁","K3sでエッジAI推論基盤を構築する際、標準KubernetesのGPUリソース管理(整数単位・request=limit固定)が壁になる。1台のGPUに複数モデルを同居させる方法とtime-slicing導入時の注意点を解説する。","2026-10-02",[10,11,12,13,14,15,16],"k3s","kubernetes","edge-ai","gpu-scheduling","device-plugin","time-slicing","fractional-gpu",1792592670708]