Re: [RFC][PATCH] split file and anonymous page queues #2

[Rik van Riel <riel@redhat.com>]

Nick Piggin wrote:
> Rik van Riel wrote:

>> We apply pressure to each of sets of the pageout queues based on:
>> - the size of each queue
>> - the fraction of recently referenced pages in each queue,
>>    not counting used-once file pages
>> - swappiness (file IO is more efficient than swap IO)

> This ignores whether a file page is mapped, doesn't it?

> Even so, it could be a good approach anyway.

It does, but once it gets the file list down to the size
where it finds that a fair number of the pages were
referenced, it will back off the pressure automatically.

Also, we do not apply the used-once algorithm to mapped
pages, meaning that mapped pages with the accessed bit
set always get rotated back onto the active list, while
unmapped pages do not.

> There are a couple of little nice improvements you have there, such as
> treating shmem pages in the same class as anon pages. We found that we
> needed something similar, so some of those things should go upstream
> on their own.

It will be hard to merge that "on its own" without the
split queues.  I can't really think of a good way to
split this patch up into multiple functional bits...

-- 
Politics is the struggle between those who want to make their country
the best in the world, and those who believe it already is.  Each group
calls the other unpatriotic.