[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v3 00/14] Series short description

On 11/18/2013 06:16 PM, Dario Faggioli wrote:
Implement vcpu soft affinity for credit1

Hello everyone,

Take 3 for his series.

Very briefly, what it does is allowing each vcpu to have:
  - an hard affinity, which they already do, and we usually call pinning. This
    is the list of pcpus where a vcpu is allowed to run;
  - a soft affinity, which this series introduces. This is the list of pcpus
    where a vcpu *prefers* to run.

Once that is done, per-vcpu NUMA-aware scheduling is easily implemented on top
of that, just by instructing libxl to issue the proper call to setup the soft
affinity of the domain's vcpus to be equal to its node-affinity.

Wrt v2 review[*] I have addressed all the comments (see individual changelogs).
In particular, I have completely redesigned the libxl interface. It now allows
both the following usage patterns:
  1. changing either soft affinity only or hard affinity only or both of them to
     the same value, and getting DEBUG or WARN output if that results in an
     inconsistent state;
  2. changing both hard and soft affinity, each one to its own value, and
     getting DEBUG or WARN output only if the *final* state is inconsistent.

The series is also available here:

  git://xenbits.xen.org/people/dariof/xen.git numa/per-vcpu-affinity-v3

Thanks and Regards,

Release-wise, I think this is probably not a blocker, but it is a very cool feature. From a scheduler standpoint, it should be fairly low risk -- it is primarily simplifying a mechanism that was introduced in 4.3. Bugs in the library code should be fairly easy to catch, and fairly low impact if they are found.

The primary thing to be concerned about is the interface; whether we have had enough time to consider the new interfaces before committing to support them. But they're fairly straightforward.

Since we're only a day past the freezing point, and (I think) almost ready to be checked in, I'm inclined to give this a freeze exception.

Any other thoughts?


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.