[Feature Request] Investigate HV Sockets for improved latency of GPU PV accelerated Hyper V VMs to host.

[JemmyBubbles]

Select Topic Area

Feature Request

Body

Background:

The looking glass project provides ultra low latency access to Windows Guests/Linux Hosts using a shared memory buffer IVSHMEM, this provides low latency / high throughput; this does not rely on network performance.

https://looking-glass.io/docs/B6/

There is no direct Equivalent to IVSHMEM in the Hyper V context, it seems as though hvsockets (for windows VMs) or vsock (linux vms) could be investigated for this.

X410 / XRPA can be utilised to provide linux gui or application streaming to the windows host over vsock. Typically this is used in the setting of WSL2 which has GPU PV support by default.

https://x410.dev/cookbook/wsl/using-x410-with-wsl2/

Suggestion:

In the setting of GPU PV accelerated windows guests it would be nice to be able to select HVsockets instead of Network to stream to the Hyper V Host. This may require the development of a custom integration service.

https://learn.microsoft.com/en-us/virtualization/hyper-v-on-windows/user-guide/make-integration-service