Jacob Tomlinson 2/7/2023

Oversubscribing GPUs in Kubernetes

Read Original

This article explains how to oversubscribe GPUs in a Kubernetes cluster using time slicing, allowing multiple Pods to share a single GPU. It covers prerequisites like setting up a Kubernetes cluster with GPUs, installing the NVIDIA Operator, and highlights caveats such as lack of memory isolation. The author notes this is suitable for development or light workloads, while recommending MIG or MPS for production.

Oversubscribing GPUs in Kubernetes

Comments

No comments yet

Be the first to share your thoughts!

Browser Extension

Get instant access to AllDevBlogs from your browser

Top of the Week