STRATUM GPU
The VM sees GPU hardware.
STRATUM sees the fabric.
The GPU no longer has to be in the server.
ACCELERATOR FABRIC.
A GPU that belongs to the datacenter.
Virtual hardware inside the VM. Physical acceleration across the Continuum.
THE MODEL.
The GPU becomes a first-class datacenter resource.
Not merely a peripheral installed in one chassis.
Virtual Hardware
Present a purpose-built PCIe accelerator device to the guest.
The guest sees BAR address windows, MMIO control, DMA-oriented transport, interrupts, telemetry, and accelerator integration as part of the VM hardware model.
Fabric
Remote acceleration can look local to the workload.
STRATUM resolves the physical accelerator, execution worker, transport, and Continuum relationship behind the virtual device.
Workload
The application should not have to understand the network beneath the GPU.
Existing Windows and Linux workloads can gain accelerator access through the VM hardware model instead of first being rebuilt as cloud-native remote-GPU applications.