Re: [RFC 2.6.11-rc2-mm2 7/7] mm: manual page migration -- sys_page_migrate

--On Wednesday, February 16, 2005 17:08:33 +0100 Andi Kleen [off-list ref] wrote:

On Wed, Feb 16, 2005 at 07:49:23AM -0800, Paul Jackson wrote:

quoted

Martin wrote:

quoted

From reading the code (not actual experiments, yet), it seems like we won't
even wake up the local kswapd until all the nodes are full. And ...

Martin - is there a Cliff Notes summary you could provide of this
subthread you and Andi are having?  I got lost somewhere along the way.

I didn't really have much thread, but as far as I understood it
Martin just wants kswapd to be a bit more aggressive in making sure
all nodes always have local memory to allocate from.

I don't see it as a pressing problem right now, but it may help
for some memory intensive workloads a bit (see numastat numa_miss output for
various nodes on how often a "wrong node" fallback happens) 

Yeah - I think I'm just worried that people are proposing a manual rather
than automatic solution to solve fallback issues. We ought to be able
to fix that without tweaking things up the wazoo by hand.

M.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>