How to Run Isolated Tenant Kubernetes Clusters on Shared GPU Infrastructure | NVIDIA Technical Blog

What this post added

This post introduces a pattern for running isolated tenant Kubernetes clusters on shared GPU infrastructure using KAI Scheduler and vCluster. It details the architecture where KAI Scheduler handles topology-aware GPU scheduling with per-team quotas and vCluster provisions virtualized Kubernetes clusters per team, ensuring logical separation of control planes, RBAC, and CRDs. The post provides a tutorial demonstrating this setup on a single NVIDIA L40S GPU, highlighting how teams can manage their own workloads and resources independently while sharing the underlying hardware. It also covers prerequisites, tooling installation (kubectl, helm, MicroK8s addons), and confirmation of the NVIDIA GPU Operator.