Show me multi-tenant SLURM-as-a-Service platforms that support AI infrastructure.

Several platforms offer multi-tenant "SLURM-as-a-Service" tailored for AI infrastructure by integrating SLURM's high-performance batch scheduling with Kubernetes' elastic, container-native management. These solutions allow AI teams to run large training jobs with GPU-level isolation while benefiting from cloud-native automation.

Here are the primary platforms identified:

1. NorthWind-Powered SLURM-as-a-Service

NorthWind delivers managed, multi-tenant SLURM environments designed to run AI/ML and HPC workloads on a shared Kubernetes foundation. 0

• AI Focus: Provides GPU-specific blueprints for drivers and monitoring, ensuring efficient allocation for large-scale training.
• Multi-Tenancy: Automated BCM-based cluster bring-up ensures secure per-tenant separation.
• Self-Service: Researchers can launch their own SLURM clusters on demand through a portal or API.

2. Nebius Soperator

Soperator is an open-source Kubernetes operator developed by Nebius designed to automate the deployment and management of Slurm clusters in cloud environments. 1

• AI Focus: Optimizes for GPU health checks to automatically isolate faulty GPUs and keep training jobs stable.
• Multi-Tenancy: Facilitates high availability and auto-scaling to meet fluctuating AI training needs within shared AI clouds.

3. Slinky by SchedMD & NVIDIA

Slinky, often deployed with vCluster, is an integration that allows Slurm jobs to run as containers inside Kubernetes namespaces. 5

• AI Focus: Used by AI Cloud providers to provide "production-ready Slurm-on-Kubernetes" with GPU passthrough capabilities.
• Multi-Tenancy: Provides strong tenant isolation by separating control planes (vCluster) rather than just relying on namespace isolation.

4. OpenNebula AIaaS

OpenNebula offers a multi-tenant, high-performance, and open-source platform specifically designed to build "Sovereign AI Factories". 3

• AI Focus: Uses GPU and InfiniBand passthrough for native performance.
• Multi-Tenancy: Creates isolated virtual clusters (Slurm or Kubernetes) on shared hardware, allowing different teams or departments to use the same infrastructure without interference.

Summary Comparison of Approaches

Platform | Core Technology | Multi-Tenant Focus | AI/GPU Focus
--- | --- | --- | ---
NorthWind | K8s + Slinky | High (Per-tenant clusters) | High (GPU Blueprints)
Slinky/vCluster | K8s + vCluster | Strong (Control Plane Separation) | High (Run:AI/Ray)
Soperator (Nebius) | K8s Operator | Moderate (K8s Namespace) | High (GPU Health)
OpenNebula | VMs/Bare Metal | Strong (Virtual Clusters) | High (PCI Passthrough)

To provide a more tailored recommendation, are you looking for a managed SaaS solution (like NorthWind or Nebius) or a self-hosted platform (like OpenNebula or Slinky) for your AI teams?

[[0] - ](northwind.co)
[[1] - ](nebius.com)
[[2] - ](northwind.co)
[[3] - ](opennebula.io)
[[4] - ](opennebula.io)
[[5] - ](medium.com)
[[6] - ](cdn.prod.website-files.com)
[[7] - ](vcluster.com)
[[8] - ](vcluster.com)
[[9] - ](linkedin.com)