AI Accelerator Fabric
AI Accelerator Fabric shows accelerator classes, available devices, active leases, and the provider resources STRATUM can assign to workloads.
AI Accelerator inventory
AI Accelerator Fabric summarizes accelerator classes, available GPUs, active leases, and the provider resources that back STRATUM GPU assignments.
A healthy inventory here confirms that the controller can see AI accelerator capacity before a VM requests it. Guest attachment and CUDA use are covered separately under the GPU & Guest Tools section in this guide.
Interface tour
AI Accelerator Fabric Inventory
Review accelerator classes, available devices, active leases, and provider resources. STRATUM can make a GPU physically located on another host, in another datacenter, at an edge site, or across a cloud-connected Continuum available to a VM as a PCI-based accelerator. The guest does not have to know where the hardware physically lives; STRATUM handles discovery, placement, transport, and presentation through the Continuum.
From the VM’s point of view, the accelerator is local. Physically, it can be anywhere in the STRATUM Continuum. Attach STRATUM GPU, install Guest Tools, and use normal CUDA/NVML applications without application-specific fabric configuration.
The outcome you should see
AI Accelerator Fabric shows the expected AI accelerator inventory and availability. VM attachment and guest use are verified under the GPU & Guest Tools section in this guide.