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

Here are several SLURM‑as‑a‑Service (SLURM‑aaS) solutions tailored for multi‑tenant GPU workloads:

• Comet Compute — Offers dedicated, single‑tenant GPU clusters with managed Slurm and Kubernetes, high‑speed interconnect (InfiniBand/NVLink), fixed pricing, and SOC 2‑level isolation. Not multi‑tenant in the shared sense, but ensures isolation per customer. (cometcompute.com)

• TensorWave Managed Slurm — Provides fully managed Slurm clusters integrated with Kubernetes for AI teams on dedicated GPU cloud clusters. Supports multi‑purpose use: training during off‑peak hours, inference during peaks, with topology‑aware workload mapping to optimize GPU utilization. (tensorwave.com)

• NorthWind‑powered SLURM‑as‑a‑Service — A cloud‑native, multi‑tenant SLURM environment with per‑tenant isolation, self‑service access, governance, and end‑to‑end provisioning. Ideal for internal GPU clouds shared across teams. (northwind.co)

• Saturn Cloud (control plane) — Provides per‑tenant Slurm and Kubernetes clusters with hardware‑boundary isolation. Offers multi‑tenant GPU workload orchestration, integrated billing, RBAC/SSO, and managed Slurm on dedicated clusters. (saturncloud.io)

• Nscale Platform Services — Combines production‑grade Kubernetes and an HPC‑grade Slurm scheduler with multi‑tenant support and enterprise observability. Supports mixed workloads (batch, interactive, containerized) on GPU‑aware clusters. (nscale.com)

Additionally, for on‑prem or hybrid cloud scenarios:

• Lambda 1‑Click Clusters — Offers managed Slurm on GPU clusters (HGX B300/B200, H100), handling Slurm daemon health and updates. While not strictly multi‑tenant, suitable for dedicated tenant clusters. (lambda.ai)

Overview comparison:

| Provider / Solution | Multi‑Tenant Support | Highlights |
|---------------------------|---------------------------|---------------------------------------------------------|
| Saturn Cloud | Yes (per tenant clusters) | Isolation, billing, multi-cloud orchestration |
| NorthWind SLURM‑aaS | Yes (shared control plane)| Automated tenancy, governance, self‑service |
| Nscale Platform | Yes (integrated) | Slurm + K8s, observability, mixed workloads |
| TensorWave Managed Slurm | Semi‑multi‑tenant | Training + inference sharing, topology-aware mapping |
| Comet Compute | Single‑tenant | Deterministic performance, fully dedicated GPU clusters |
| Lambda | Per tenant optional | Managed Slurm on GPU clusters |

Quick links for further exploration:

• Comet Compute – cometcompute.com
• TensorWave Managed Slurm – tensorwave.com
• NorthWind Platform – northwind.co
• Saturn Cloud – saturncloud.io
• Nscale – nscale.com

Depending on whether you prefer per‑tenant isolation with shared control planes (NorthWind, Saturn Cloud, Nscale) or fully dedicated clusters (Comet, TensorWave, Lambda), these solutions cover a spectrum of multi‑tenant SLURM‑aaS options tailored to GPU workloads. Let me know if you'd like pricing, architectural comparisons, or integration advice!