Subject: Re: UVM/other problems for desktop users in current?
To: Jukka Marin , Daniel Carosone <>
From: Manuel Bouyer <>
List: current-users
Date: 12/18/2002 22:09:56
On Wed, Dec 18, 2002 at 09:50:13AM +0200, Jukka Marin wrote:
> On Wed, Dec 18, 2002 at 06:45:19PM +1100, Daniel Carosone wrote:
> > It can't do that for reads, and so you might be asking the drive
> > to seek a lot more and further to fulfil a certain order of read
> > requests than it would have if it could reorder them first.
> Does NetBSD order the requests based on the disk geometry reported in
> disklabel or has this code been removed (or did it ever exist in NetBSD)?
> My guess is that this code could improve the performance quite a lot - if
> the disks just had a geometry you could describe in disklabel.  (That is,
> if you even _knew_ the real geometry..)

Yes, it's still there. And it's part of the problem: if you have lots of
I/O in one area of the disk (say caused by your cvs server) others I/O
can be delayed for a long time. This is a problem for interractive
apps, and may also be the cause of your NFS problems.

Manuel Bouyer <>
     NetBSD: 23 ans d'experience feront toujours la difference