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

The clearest GPU-first platforms in your results that explicitly offer or enable Slurm-as-a-Service are NorthWind GPU PaaS and Crusoe Managed Slurm on CMK.[1][8] A third relevant option is AMD Spur / Spur-Cloud, which is a GPU-first scheduler with Slurm-compatible interfaces and a cloud platform layer, but it is not literally Slurm.[6]

• NorthWind GPU PaaS: NorthWind says its GPU PaaS plus Project Slinky lets enterprises and cloud providers deliver “Slurm-as-a-Service on Kubernetes” with self-service, multi-tenant clusters and GPU access.[1]
• Crusoe Managed Slurm on CMK: Crusoe describes a managed Slurm offering on its Managed Kubernetes platform, aimed at single-tenant GPU training clusters with a familiar Slurm interface.[8]
• NVIDIA Slinky stack: NVIDIA/SchedMD describe Slinky’s slurm-operator approach for running full Slurm clusters on Kubernetes, which is the underlying mechanism many “Slurm-as-a-Service” offerings use.[2][9]
• AMD Spur-Cloud: Spur is a GPU-first job scheduler with Slurm-compatible CLI/API support, and Spur-Cloud extends it into a GPU-as-a-Service platform with web UI, billing, and multi-tenancy.[6]

If you want the most direct match to “SLURM-as-a-Service,” NorthWind is the strongest explicit fit.[1] If you want a managed Slurm platform for GPU training, Crusoe is the other clear match.[8]

Source URLs:

• northwind.co
• crusoe.ai
• developer.nvidia.com
• rocm.blogs.amd.com