Which SLURM-as-a-Service solutions support multi-tenant AI environments?

Below are SLURM‑as‑a‑Service (SLURMaaS) offerings that explicitly support multi‑tenant AI environments (isolation, per‑team clusters, identity controls, or namespace separation).

1) NorthWind – NorthWind‑Powered SLURM‑as‑a‑Service

NorthWind delivers fully managed, on‑demand SLURM clusters with secure per‑tenant separation and governance. It’s designed for providers and enterprises that need standardized, isolated SLURM environments for AI/ML and HPC.

Key capabilities:

• Self‑service SLURM cluster provisioning
• Multi‑tenant isolation
• Works across bare metal, VMs, and Kubernetes

Source:
northwind.co (northwind.co)
NorthWind deployment options (notes multi‑tenant AI/GPU environments):
docs.northwind.co (docs.northwind.co)

Best for: Enterprises or platform providers building internal AI clouds with strong tenant separation.

2) CoreWeave – SUNK (Slurm on Kubernetes)

CoreWeave’s SUNK (Slurm on Kubernetes) is a managed SLURM environment purpose‑built for AI training at scale. It preserves SLURM workflows while leveraging Kubernetes for cluster operations and identity.

Multi‑tenant–relevant features:

• Unified Slurm + Kubernetes operations
• Automated user provisioning & identity management
• Production‑grade lifecycle control
• Available on CoreWeave or via SUNK Anywhere (multi‑cloud)

Sources:
Product overview: coreweave.com (coreweave.com)
Docs (SUNK description): docs.coreweave.com (docs.coreweave.com)
SUNK Anywhere (multi‑cloud support): coreweave.com (coreweave.com)

Best for: AI‑native companies running large multi‑GPU training across teams with shared GPU infrastructure.

3) Google Cloud – Cluster Director (Managed Slurm)

Google Cloud’s Cluster Director provides a managed infrastructure service for Slurm and Kubernetes, including AI/ML training templates. It supports enterprise IAM and GCP project‑level isolation for multi‑team environments.

Source:
cloud.google.com (cloud.google.com)

Best for: Organizations already standardized on Google Cloud needing managed Slurm integrated with GCP IAM, networking, and GPU instances.

Quick Comparison

| Vendor | Multi‑Tenant Model | Managed? | Multi‑Cloud? |
|---------|-------------------|----------|--------------|
| NorthWind | Per‑tenant SLURM clusters | Yes | Yes |
| CoreWeave SUNK | Slurm-on-K8s w/ identity controls | Yes | Yes (SUNK Anywhere) |
| Google Cluster Director | GCP IAM/project isolation | Yes | No (GCP) |

If you share your deployment target (on‑prem, hybrid, AWS/GCP/Azure) and whether you need hard isolation vs namespace isolation, I can narrow this further.