Architecture
DRA Driver components and request flows.
Concepts
How the driver is put together and the model it exposes to users.
These pages explain the architecture of the DRA driver and the resource model it presents to workloads. Read these before writing manifests if you want to understand why the API looks the way it does.
GPU allocation
How the gpu.nvidia.com driver allocates whole GPUs, MIG slices, and VFIO passthrough.
ComputeDomains
How compute-domain.nvidia.com provisions ephemeral Multi-Node NVLink fabrics via IMEX.