Project Lightning Talk: K8s issue #52757: Sharing GPUs Among Multiple Containers
Lightningat KubeCon + CloudNativeCon China 2025 · on Tuesday, 2025-06-10· at 11:28· Level 16 | Grand Ballroom I·
summary
Covers the longstanding Kubernetes issue #52757 about flexibly sharing GPUs among multiple containers, especially for short-lived AI inference workloads. Describes solutions and practices using Dynamic Resource Allocation (DRA) and the CNCF sandbox project HAMi, discusses challenges of GPU sharing, approaches for sharing non‑NVIDIA AI chips, and how these sharing technologies integrate with Volcano, Koordinator, and Kueue.
AI-written summary · about
related talks
34mDeclarative Device Virtualization: Orchestrating GPUs & Hardware in Cloud Native Environments
16mGPU Accelerated Containers on Apple Silicon with libkrun and podman machine
25mRunning I/O Intensive Workloads on Kubernetes
10mLightning Talk
30mIntroduction and Deep Dive into containerd
22mEmbedded Software Development Efficiency Improvement with Docker and Kubernetes
31mYour Containers Aren’t Alone: Demystifying Container Isolation
14mYour Containers Aren’t Alone: Demystifying Container Isolation