The task
Use GPU capacity more flexibly than a fixed server-to-accelerator pairing allows.
Dutch public-sector organisation
GPU capacity that follows the workload.
A GPU cluster where compute can be assigned to hosts through a PCIe fabric, with Slurm for workload scheduling.
What was built
A composable setup with RTX 8000 GPUs, a PCIe fabric and Slurm.
The scope
The project describes infrastructure and GPU assignment. It makes no unsubstantiated utilisation, cost or speed claims.
VF / 01REFERENCE ARCHITECTURE
- People and applications: CHAT / API / DEVELOPER TOOLS
- Access · API gateway: IDENTITY / LIMITS / ROUTING
- AI agents, retrieval and models: TOOLS / SOURCES / MODELS
- GPU · Storage · MLOps: KUBERNETES / NVMe-oF / OBSERVABILITY
REFERENCE, NOT A NETWORK MAP
The pattern behind the solution.
A generic overview of the building blocks. Technical addresses, internal systems and client identities are excluded from publication.
Explore the architecture ↗