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

For SLURM-as-a-Service aimed at multi-tenant GPU workloads, the clearest solutions in your results are NorthWind GPU PaaS + Project Slinky and RedFort Tech Managed SLURM. NorthWind explicitly describes a secure, multi-tenant, self-service model where users can provision their own Slurm cluster inside a namespace on shared Kubernetes infrastructure.[1] RedFort Tech markets a managed SLURM-as-a-Service offering with preconfigured controller/login nodes and managed GPU compute nodes.[4]

A practical shortlist:

| Solution | Why it fits multi-tenant GPU SLURM | Notes |
|---|---|---|
| NorthWind GPU PaaS + Project Slinky | Self-service Slurm clusters on Kubernetes; isolated per-user/namespace; shared GPU infra | Strongest match for “as-a-service” multi-tenancy.[1] |
| RedFort Tech Managed SLURM | Managed SLURM-as-a-Service with GPU nodes and automated orchestration | More traditional managed HPC delivery.[4] |
| Oracle OCI on-demand HPC/AI platform | Multi-tenant GPU/CPU access, Slurm orchestration, autoscaling, quotas/chargeback | Broader platform play, not SLURM-only.[11] |
| Slurm vService (CSCS / SchedMD talk) | Multi-tenant service model with Kubernetes orchestrating Slurm daemons | More of a reference architecture than a product.[14] |

If your priority is platform-native self-service for multiple tenants, NorthWind is the most directly aligned option because it combines Kubernetes tenancy boundaries with Slurm cluster provisioning per user/team.[1][8] If you want a managed HPC provider that will run Slurm for you, RedFort is the more straightforward fit.[4]

For background on GPU multi-tenancy techniques that often accompany these offerings, Slurm supports GPU GRES and MIG-style partitioning, which can help isolate tenants and improve utilization.[10][2][9]

Source URLs:

• northwind.co
• redforttech.us
• docs.oracle.com
• slurm.schedmd.com