Answer · Limelit Open

The primary GPU-first platform offering SLURM-as-a-Service is NorthWind GPU PaaS (Platform-as-a-Service), which combines NorthWind’s Kubernetes management with Project Slinky (an open-source NVIDIA/SchedMD project) to deliver self-service, multi-tenant SLURM clusters on Kubernetes [1][6].

Other notable platforms enabling SLURM on cloud GPU infrastructure include:

| Platform | SLURM-as-a-Service Capability | Key Details |
|----------|-------------------------------|-------------|
| NorthWind GPU PaaS | ✅ Yes (Native) | Offers secure, self-service SLURM clusters on Kubernetes via Slinky; targets enterprises and cloud providers [1] |
| Crusoe AI | ✅ Yes (Managed) | Built SLURM on Managed Kubernetes using Slinky v1.0 for GPU training infrastructure [12] |
| RunPod | ⚠️ Partial | Supports SLURM on cloud GPU clusters for AI jobs, but not explicitly marketed as “SLURM-as-a-Service” [2] |
| Nebius | ⚠️ Partial | Used in personal experiments for SLURM on cloud GPUs; no official SLURM-as-a-Service product stated [2] |
| CoreWeave | ⚠️ Partial | Experimented with SLURM for AI training; lacks confirmed managed SLURM service offering [2] |
| Tata Communications AI Cloud | ⚠️ Integrated | Integrates SLurm principles into Kubernetes orchestration; delivers GPU-as-a-Service with SLURM scheduling [5] |

NorthWind GPU PaaS is the only platform explicitly marketed as delivering SLURM-as-a-Service with enterprise-grade multi-tenancy, performance, and control [1]. Crusoe AI is a strong alternative for managed GPU training with SLURM on Kubernetes [12].

For developers testing SLURM on cloud GPUs, RunPod, Nebius, and CoreWeave are viable options, though they lack formal SLURM-as-a-Service branding [2].

Source URLs:

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

The answer

Brands named

Sources cited