Re: [Xen-devel] 4.2 Release Plan / TODO

[Dario Faggioli <raistlin@xxxxxxxx>]

On Tue, 2012-04-24 at 13:53 +0100, Ian Campbell wrote: 
> Plan for a 4.2 release:
> http://lists.xen.org/archives/html/xen-devel/2012-03/msg00793.html
> 
> The time line is as follows:
> 
> 19 March        -- TODO list locked down
> 2 April         -- Feature Freeze
>                                               << WE ARE HERE
>
I'm guessing we're still here and hence I'm proposing the following.

> The updated TODO list follows. Per the release plan a strong case will
> now have to be made as to why new items should be added to the list,
> especially as a blocker, rather than deferred to 4.3.
> 
As it is probably known and has been repeatedly stated in various
threads, xm/xend has some kind of NUMA placement policy when a new guest
is being created. On the opposite hand, xl/libxl does absolutely
nothing, and that looks like an issue for the xl-xend feature parity
we're aiming at.

What xend basically does is trying to figure out where to put a guest
(according to some heuristics) and it then pins its vcpus there. The
NUMA series I posted also introduces some placement heuristics into xl,
and I can easily take the scheduling/affinity modification apart and do
pin vcpus basing on the outcome of the heuristics, just like xend.

Yes, I noticed this for the first time way earlier than now, but it's
only now that it came to my mind that this could be quite a serious
feature parity issue, and that I can (try to!) fix with a reasonably
small effort.

Therefore, I'm officially proposing to add "automatic NUMA placement and
vcpu pinning in xl" here below:

> tools, blockers:
> ... 
> * xl compatibility with xm:
>               * [BUG] cannot create an empty CD-ROM driver on HVM guest,
>                 reported by Fabio Fantoni in
>                 <4F9672DD.2080902@xxxxxxxxxx>
>               * [BUG] does not honour scheduler weight params, reported
>                 by Dieter Bloms in <20120423193518.GA16134@xxxxxxxx>,
>                 Dieter has posted a patch.
                * does not automatically try to select a (set of) 
                  node(s) on which to create a VM and pin its vcpus
                  there.

The strong case I'm making is that not having this is a regression wrt
default xend behaviour, and it could upset and break stuff for people
expecting the toolstack tacking care about NUMA, at least in some rough
way (which btw is what xend is doing).

I'm not far from having the patches so, if the proposal is accepted, I
can post them next week, I'm quite sure.

Thoughts?

Thanks and Regards,
Dario

-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://retis.sssup.it/people/faggioli
Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel