SignalTech/AI
Memory-Enforced GPU Sharing in Kubernetes with NVIDIA CUDA 13.4
High-priority alert based on source significance and keyword relevance.
Context
What happened: NVIDIA CUDA 13.4 adds a memory enforcement boundary with a new time-sharing mode, alongside the existing sm-sharing compute mode in MPS. A new open-source Kubernetes operator, gpu-fractioning, is hosted under the KAI Scheduler organization on GitHub.
Key numbers: KAI Scheduler v0.18.0 introduces a GPU-sharing mode called
Why it matters: The operator applies per-container memory limits and compute mode on each GPU node, and KAI Scheduler is the first scheduler integrated with it. Any Kubernetes scheduler that can write the annotations can enforce GPU fractions with CUDA.
What changed
NVIDIA CUDA 13.4 adds a memory enforcement boundary with a new time-sharing mode, alongside the existing sm-sharing compute mode in MPS. A new open-source Kubernetes operator, gpu-fractioning, is hosted under the KAI Scheduler organization on GitHub.
Why it matters
The operator applies per-container memory limits and compute mode on each GPU node, and KAI Scheduler is the first scheduler integrated with it. Any Kubernetes scheduler that can write the annotations can enforce GPU fractions with CUDA.