Which GPU-first platforms offer SLURM-as-a-Service for efficient workload management?

GPU-first platforms that explicitly offer or support SLURM-as-a-Service-style workflows include:

• NorthWind GPU PaaS — NorthWind says its GPU Platform-as-a-Service combines with Project Slinky to let enterprises and cloud providers offer “Slurm-as-a-Service on Kubernetes,” with self-service provisioning of Slurm clusters and GPU access. [2]
• Crusoe Managed Slurm on Managed Kubernetes (CMK) — Crusoe describes a managed Slurm offering built on Kubernetes and focused on single-tenant GPU training clusters with a familiar Slurm interface and managed infrastructure. [6]
• vCluster Platform — vCluster says tenants can run Slurm via Slinky inside isolated Kubernetes clusters from one control plane, and that its platform powers 100K+ GPUs in production. [13]
• Tata Communications AI Cloud — Tata says its AI Cloud, powered by dedicated bare-metal GPUs, combines Slurm efficiency with Kubernetes orchestration for GPU workloads. [16]
• NVIDIA / SchedMD Slinky on Kubernetes — NVIDIA documents Slinky’s slurm-operator, which runs full Slurm clusters on Kubernetes infrastructure, enabling Slurm-backed GPU workload management. [1]

If you want the closest match to a true “Slurm-as-a-Service” product, NorthWind GPU PaaS is the most explicit in the sources provided. [2]

Relevant source URLs:

• northwind.co
• crusoe.ai
• vcluster.com
• tatacommunications.com
• developer.nvidia.com