Re: [RFC 5/8] Make writeout during reclaim cpuset aware

[Andi Kleen]

On Wednesday 17 January 2007 15:20, Paul Jackson wrote:
> Andi wrote:
> > Is there a reason this can't be just done by node, ignoring the cpusets?
>
> This suggestion doesn't make a whole lot of sense to me.
>
> We're looking to see if a task has dirtied most of the
> pages in the nodes it is allowed to use.  If it has, then
> we want to start pushing pages to the disk harder, and
> slowing down the tasks writes.
>
> What would it mean to do this per-node?  And why would
> that be better?

With a per node dirty limit you would get essentially the
same effect and it would have the advantage of helping
people who don't configure any cpusets but run on a NUMA 
system.

-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/