Re: [PATCH 08 of 10 v3] libxl: enable automatic placement of guests on NUMA nodes

[Dario Faggioli <raistlin@linux.it>]

[-- Attachment #1.1: Type: text/plain, Size: 1652 bytes --]

On Fri, 2012-07-06 at 14:21 +0100, Ian Campbell wrote: 
> > Ok, I see. What about something like the below:
> > 
> > "However, if the amount of free memory of two candidates is very
> > similar, we look at how many domains are assigned to each candidate, and
> > take the one that has fewer of them."
> 
> Looks good, thanks.
> 
Ok.

> Is it still "very similar" now that it is a total order or is it
> actually a strict equality now? (or an equality down to the precision of
> a float...)
> 
It is, at least according to me. In fact, if the normalized difference
between the amount of free memory of two candidates is 0.1 (i.e.,
candidates are "very similar") and the difference in number of domains
is -0.5 (one has half the domains the other does) we get
3*0.1-0.5=-0.2 ==> the number of domains is important. OTOH, a
normalized difference of 0.2 in the amount of free memory (i.e.,
candidates are no longer "very similar") is enough to prevent the same
difference in number of domains from subverting the sign of the
comparison.

It of course depends on how much details we want to go in in this kind
of documentation. Given there's no way neither an application developer
nor an user could change/affect the parameters controlling the
heuristics, I think what we have there is just fine.

Did I persuade you? :-P

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 #1.2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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