Name of function: vGPU support User story: As a cloud user. I need to be able to access GPU resources when I need them for complex rendering and other tasks so I can complete complex tasks quickly. As an administrator, I want to be able to offer users vGPU capabilities by sharing a physical networking card between multiple guests to save the expense of lots of hardware deployments. Purpose of function: Direct access to a vGPU allows users to share one card for multiple instances, using only one physical GPU. Sharing hardware to save resources. Working on director to automate deployment. Based on drivers from NVIDIA. Work required (from detail plan): Procedural ------ How to creating a custom image (specific change - remove a driver “nouveau” How to configure Nova How to check proper install & config: run the command to check. CLI with NVIDIA command ------ Conceptual ------ Need a licence for driver (GRID virtual PC or Quadro vDWS) What are the limitations of the driver (one type of GPU) List of limitations (in presentation) List of supported hardware Sizing of cards - specific matrix on what you can and can’t mix (make doc so that other types of GPUs can be included in the future.) Caveats based on devices chosen (may not be long term) ------ Reference ------ N/A ------
Adding the estimated hours to complete - numbers provided by Mikey in the detail plan.
Accepting into the RHOSP 14 documentation program, and assigning to Mikey for review.
Peer review completed, content published to OSP14: https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/14/html-single/instances_and_images_guide/index#ch-virtual-gpu