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

Re: [Xen-devel] [PATCH 04 of 11 v3] xen: sched_credit: let the scheduler know about node-affinity

On Fri, 2013-02-01 at 11:01 +0000, Dario Faggioli wrote:
> As vcpu-affinity tells where VCPUs must run, node-affinity tells
> where they should or, better, prefer.

I cannot parse this sentence. Is the "or, better," redundant?

>  While respecting vcpu-affinity
> remains mandatory, node-affinity is not that strict, it only expresses
> a preference, although honouring it is almost always true that will
> bring significant performances benefit

I can't parse this last bit either ("although...benefit"). If I drop the
"is almost always true that" it makes sense and I think expresses what
you meant. And just a nit, it is "performance benefits".

>  (especially as compared to
> not having any affinity at all).
> This change modifies the VCPU load balancing algorithm (for the
> credit scheduler only), introducing a two steps logic.
> During the first step, we use the node-affinity mask. The aim is
> giving precedence to the CPUs where it is known to be preferable
> for the domain to run. If that fails in finding a valid PCPU, the
> node-affinity is just ignored and, in the second step, we fall
> back to using cpu-affinity only.

I think from previous paragraphs that you mean to say that the first
path takes into account node and vcpu affinity while the second only
vcpu affinity? The above reads as if it only uses the node affinity on
the first pass.

I'm not really well placed to review the actual meat, but I see George
has some comments.


Xen-devel mailing list



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