How to Configure Multi-Tenant GPU Scheduling with NVIDIA Run:ai

This article provides a comprehensive tutorial on configuring multi-tenant GPU scheduling with NVIDIA Run:ai in a Kubernetes cluster. It explains how to transform a shared GPU cluster into a governed multi-tenant platform by organizing workloads into departments and projects, assigning guaranteed GPU quotas per node pool, and allowing controlled over-quota usage. The design relies on four key controls: Quota (resource entitlement), Fairshare (distribution of unused capacity), Priority (workload ordering), and Preemptibility (reclaiming borrowed capacity). It covers mapping organizational boundaries, setting up node pools, defining workload policies, and separating production from development scheduling. The goal is predictable, measurable, and recoverable GPU allocation aligned with business ownership.