Find SLURM-as-a-Service solutions for multi-tenant GPU workloads.

If you want SLURM-as-a-Service for multi-tenant GPU workloads, the clearest match in the results is NorthWind-Powered SLURM-as-a-Service. It is described as a fully managed, multi-tenant SLURM offering with self-service cluster launch, per-tenant isolation, and governance built in.[1]

Other relevant options and patterns:

• NorthWind GPU PaaS + Slinky: positioned as a way to provide secure, multi-tenant, self-service Slurm clusters on shared Kubernetes infrastructure.[2][9]
• Oracle OCI reference architecture: an on-demand HPC and AI platform using Open OnDemand + Slurm + GPU autoscaling with multi-tenant GPU/CPU access; this is more of a reference deployment than a managed SaaS product.[12]
• NVIDIA DGX Cloud Slurm: supports managed Slurm clusters for GPU workloads, but the result is more about cluster usage/examples than explicit multi-tenancy as a service.[16]
• Crusoe Managed Slurm on CMK: managed Slurm on Kubernetes, but explicitly described as single-tenant, so it does not fit your multi-tenant requirement.[17]

For the strongest fit, shortlist:

• NorthWind-Powered SLURM-as-a-Service[1]
• NorthWind GPU PaaS with Slinky[2][9]
• OCI on-demand HPC/AI platform if you want a deployable architecture rather than a product[12]

Useful product/source pages:

• northwind.co
• northwind.co
• docs.oracle.com
• developer.nvidia.com
• crusoe.ai