Subject: Re: Bad response...
To: Eric Haszlakiewicz <erh@jodi.nimenees.com>
From: Pavel Cahyna <pavel.cahyna@st.cuni.cz>
List: current-users
Date: 08/31/2004 18:08:39
On Mon, 30 Aug 2004 23:23:32 +0000, Eric Haszlakiewicz wrote:

> 	The problem is that the system had previously written out pages of
> that interactive program because the LRU algorithm thought they were 
> unused and that it had a whole bunch of higher priority file data to
> operate on.

I thinks similar problems were in linux 2.2 . See a message from FreeBSD's
Matt Dillon to linux developers, where he describes the algorithms used by
FreeBSD. http://mail.nl.linux.org/linux-mm/2000-05/msg00419.html

IIRC linux 2.4 tried to adopt those suggestions, but finally a simpler
solution was adopted.

See:
ftp://ftp.suse.com//pub/people/andrea/talks/english/2001/pluto-dec-pub-0.tar.gz
(file pluto.ps)

and
http://www.csn.ul.ie/~mel/projects/vm/guide/html/understand/
especially section 11.6

> A solution would be to give those interactive program memory pages
> a higher prority than the file pages so the file pages get
> considered flushable first.  (how to do this?  I have no idea. :) )

Maybe it would be sufficient to give pages referenced multiple times
higher priority? That way, pages read or written sequentially would
not fill the cache, since they would be evicted first.

Bye	Pavel