Open Stack

> From: Jiang, Yunhong [mailto:yunhong.jiang at intel.com]
> > From: John Garbutt [mailto:John.Garbutt at citrix.com]
> > 1) What are the use cases for the more general pass-through?
> > I understand the "want GPU flavour" concept, for desktops or HPC
> workloads.
> 
> SR-IOV is another popular reason. For example, SR-IOV NIC pass-through so
> that Instance can utilize the network card directly, also some platform has
> SR-IOV encryption card etc.
> 
> Another one is InfiniBand, although I'm not sure the latest progress in VMM
> side. In 2007, there are some support for IniniBand PV support in Xen
> community.
> > Is the idea for higher performance do the device pass-through?

I was more meaning why would people want SR-IOV, my assumption is because of better performance.
Certainly encryption on the card is an interesting extra example I had not considered.

> > Would the extra spec try to say: "high performance local disk"?
> 
> I think it depends on how will we implement the interface. For NIC, user
> may select 1G NIC card, 10G NIC card. But no idea of other devices, like , how
> do you think should GPU with different capability be presented to user?

I assumed admins would group "similar" hardware with host-aggregates rather than assume exact matches of GPUs. And the spec would relate to the host aggregate, and suggest "GPU pass-through", rather than "GPU pass-through of NVidia SuperDooper42"
 
Thanks for the extra info. That seems to have cleared things up a bit.

John