From: Jeremy Fitzhardinge <jeremy@goop.org>
To: Andrew Morton <akpm@zip.com.au>
Cc: Constantin Loizides <Constantin.Loizides@isg.de>,
Linux Kernel List <linux-kernel@vger.kernel.org>
Subject: Re: [Ext2-devel] disk throughput
Date: 09 Nov 2001 00:49:17 -0800 [thread overview]
Message-ID: <1005295757.5672.4.camel@ixodes.goop.org> (raw)
In-Reply-To: <3BEB72DA.7ADA8332@zip.com.au>
In-Reply-To: <3BEAA3B7.A7F9328E@isg.de>, <3BE647F4.AD576FF2@zip.com.au> <Pine.GSO.4.21.0111050904000.23204-100000@weyl.math.psu.edu> <3BE71131.59BA0CFC@zip.com.au> <20011106105138Z16653-12382+40@humbolt.nl.linux.org> <1005063444.25686.18.camel@ixodes.goop.org> <3BEAA3B7.A7F9328E@isg.de> <1005237961.4537.0.camel@ixodes.goop.org> <3BEB72DA.7ADA8332@zip.com.au>
On Thu, 2001-11-08 at 22:08, Andrew Morton wrote:
> How do you avoid aftifacts from Linux caching with this test?
Randomly seeking around a 40G disk isn't going to give the caching much
to work with (that was on a 256Mbyte machine). Also, I tried it with
raw devices, but it made no difference to the measurements.
> > I found the random seeking was the only way I could get reasonable
> > numbers out of the drive; any of the ordered patterns of the form "OK,
> > lets measure N block seeks" seemed to be predicted by the drive and gave
> > unreasonably fast results.
> >
>
> But we *want* unreasonably fast results! We need to understand
> this device-level prediction, then decide if it's sufficiently
> widespread for us to go and exploit the hell out of it.
>
> Have you any ideas as to what's happening?
Not really. I wanted results which at least confirm the simple model of
how the disk works. When I did deterministic seeks, it was something
like:
for(n = 0; n < disksize/2; n += about_a_cylinder) {
seek(n);
read();
seek(disksize-n);
read();
}
I suspect that the drive was partitioning its cache and using lots of
readahead to satisfy the reads at either end of the seek. I was trying
to work around that by making the n increment large enough, but it
didn't seem to work. Maybe the drive was being extra special clever, or
maybe I miscalculated the cylinder size. Or maybe I just happened to be
hitting a rotational sweet-spot.
For this kind of test, it would be interesting to know what differences
there are between read and write. In principle the drive could start
writing a track anywhere it wants so long as it has enough to write the
whole thing. That would automatically make it get the track-track skew
right...
J
next prev parent reply other threads:[~2001-11-09 8:49 UTC|newest]
Thread overview: 79+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-11-05 2:13 disk throughput Andrew Morton
2001-11-05 3:20 ` Mohammad A. Haque
2001-11-05 3:31 ` Andrew Morton
2001-11-05 3:32 ` [Ext2-devel] " Mike Fedyk
2001-11-05 3:45 ` Andrew Morton
2001-11-05 4:39 ` Mike Fedyk
2001-11-05 7:06 ` Jens Axboe
2001-11-05 7:14 ` Andrew Morton
2001-11-05 7:26 ` Jens Axboe
2001-11-05 7:14 ` Mike Fedyk
2001-11-05 7:18 ` Jens Axboe
2001-11-05 7:18 ` Jens Axboe
2001-11-05 9:14 ` Mike Fedyk
2001-11-05 9:20 ` Jens Axboe
2001-11-05 5:54 ` Albert D. Cahalan
2001-11-05 8:04 ` Andrew Morton
2001-11-05 12:28 ` Matthias Andree
2001-11-05 14:23 ` Alexander Viro
2001-11-05 22:22 ` Andrew Morton
2001-11-05 22:41 ` Andreas Dilger
2001-11-05 22:53 ` Andrew Morton
2001-11-08 15:28 ` Constantin Loizides
2001-11-05 23:14 ` Dan Hollis
2001-11-06 10:52 ` Daniel Phillips
2001-11-06 16:17 ` Jeremy Fitzhardinge
2001-11-08 15:24 ` Constantin Loizides
2001-11-08 16:46 ` Jeremy Fitzhardinge
2001-11-09 6:08 ` Andrew Morton
2001-11-09 8:49 ` Jeremy Fitzhardinge [this message]
2001-11-06 21:45 ` Stephen Tweedie
2001-11-05 20:16 ` Andreas Dilger
2001-11-05 20:28 ` m
2001-11-05 21:39 ` Andrew Morton
2001-11-05 22:59 ` Linus Torvalds
2001-11-05 23:36 ` Alexander Viro
2001-11-05 23:50 ` Linus Torvalds
2001-11-06 0:03 ` Linus Torvalds
2001-11-06 1:33 ` Alexander Viro
2001-11-06 2:10 ` Linus Torvalds
2001-11-06 3:02 ` Alexander Viro
2001-11-06 8:39 ` Alan Cox
2001-11-06 8:37 ` Alexander Viro
2001-11-06 8:48 ` Andrew Morton
2001-11-06 3:49 ` Alexander Viro
2001-11-06 4:01 ` Linus Torvalds
2001-11-06 4:21 ` Alexander Viro
2001-11-06 5:01 ` Linus Torvalds
2001-11-06 5:31 ` Andrew Morton
2001-11-06 5:48 ` Linus Torvalds
2001-11-06 7:34 ` Mike Castle
2001-11-06 7:10 ` Kai Henningsen
2001-11-09 22:35 ` Riley Williams
2001-11-06 1:28 ` Alexander Viro
2001-11-06 9:16 ` Wojtek Pilorz
2001-11-06 9:58 ` Alexander Viro
2001-11-08 12:51 ` Pavel Machek
2001-11-06 21:48 ` Stephen Tweedie
2001-11-06 23:17 ` ext2/ialloc.c cleanup Alexander Viro
2001-11-07 19:34 ` [Ext2-devel] " Andreas Dilger
2001-11-07 20:02 ` Alexander Viro
2001-11-08 2:06 ` Andrew Morton
2001-11-08 20:45 ` Andrew Morton
2001-11-08 22:16 ` Alexander Viro
2001-11-08 22:43 ` Andreas Dilger
2001-11-08 23:08 ` Alexander Viro
2001-11-09 6:15 ` Andrew Morton
2001-11-09 6:56 ` Andreas Dilger
2001-11-09 7:09 ` Andrew Morton
2001-11-09 7:12 ` Alexander Viro
2001-11-09 7:18 ` Andrew Morton
2001-11-05 9:45 ` [Ext2-devel] disk throughput Alex Bligh - linux-kernel
2001-11-05 9:58 ` Alex Bligh - linux-kernel
2001-11-05 8:47 ` Jan Kara
2001-11-05 8:50 ` [Ext2-devel] " Mike Fedyk
2001-11-05 9:01 ` Jan Kara
2001-11-05 12:23 ` Matthias Andree
2001-11-05 22:39 ` Andrew Morton
2001-11-05 23:41 ` Matthias Andree
2001-11-12 6:04 [Ext2-devel] " Yan, Noah
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1005295757.5672.4.camel@ixodes.goop.org \
--to=jeremy@goop.org \
--cc=Constantin.Loizides@isg.de \
--cc=akpm@zip.com.au \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).