Open Stack

On 07/08/2016 09:59 AM, Brent Troge wrote:
> context - high performance private cloud with cpu pinning
>
> Is it possible to map vCPUs to specific pCPUs ?
> Currently I see you can only direct which vCPUs are mapped to a specific
> NUMA node
>
> hw:numa_cpus.0=1,2,3,4
>
> However, to get even more granular, is it possible to create a flavor
> which maps vCPU to specific pCPU within a numa node ?
>
> Something like:
> hw:numa_cpus.<NUMA-NODE>-<pCPU>=<vCPU>
>
> hw:numa_cpus.0-1=1
> hw:numa_cpus.0-2=2

I presume you have more than a single compute node in your deployment?

If you had a system that had a flavor pin vCPUs to individual pCPUs, you 
would essentially be stating that an instance launched with that flavor 
would always consume that particular pCPU set.

In that case, what you have isn't a cloud, which is defined by a 
virtualized, API-driven, hardware-abstracted, service-driven system; 
what you have is just a specialized hardware appliance over which you 
are layering a REST API that wasn't designed for that appliance.

The NUMA placement policy controls that exist in Nova already allow a 
(sometimes ludicrous) amount of hardware-specific control for what 
should be a virtualized abstraction of that hardware. I personally don't 
care to see our abstraction layer any further destroyed with 
hardware-specific interfaces.

Best,
-jay