From: Hans Reiser <reiser@namesys.com>
To: Chris Mason <mason@suse.com>
Cc: bernz@bernztech.org, reiserfs-list@namesys.com
Subject: Re: slow stat() ?
Date: Sun, 30 Nov 2003 23:47:00 -0800 [thread overview]
Message-ID: <3FCAF1F4.6040106@namesys.com> (raw)
In-Reply-To: <1070299845.12809.98.camel@tiny.suse.com>
Chris Mason wrote:
>On Mon, 2003-12-01 at 12:21, David Bernick wrote:
>
>
>>Hi all,
>>
>>First, what I'm running Reiser on:
>>
>>Supermicro dual Xeon motherboard with 2x2.6Ghz xeons in hyperthreaded
>>mode. 3ware 7500-8 (or whatever the call the 8 port ATA card) with 8x250GB
>>disks attatched. 4GB RAM.
>>
>>Software:
>>RedHat 9 with latest up2date
>>Kernel 2.4.22 no patches
>>The journal is not on a separate set of spindles.
>>
>>/proc/sys/vm/max-readahead=512
>>/proc/sys/vm/min-readahead=128
>>/proc/sys/vm/bdflush
>>100 5000 640 2560 150 30000 60 20 0
>>
>>Okay. Now for the issue at hand. When the logical disk is being read from,
>>reiser is fast and effecient. We have millions (10M?) of files (about 1M
>>each) in thousands of directories. Reiser reads well (exporting the disk
>>over NFS) and we're very impressed by it. It's a bit slow on the writes
>>and the system load goes up marginally, but that's not a big deal. The
>>thing that puzzles me is that when we perform functions like "find" or in
>>perl "-e filename" or anything that uses fstat() or stat64() or any of
>>those functions brings the system usage to 100% and the load climbs
>>quickly.
>>
>>Are there any patches, kernel updates, tuning, etc that can help with this
>>problem?
>>
>>
>
>You could boot with profile=2 and send the results of a readprofile
>command taken after the system load goes very high. This will slow down
>your box for general usage, so you would only want to run with profile=2
>for a short time.
>
>My guess is the stats are being done across a large directory tree and
>this is generating lots of atime updates. You could mount -o
>noatime,nodiratime, which will reduce the amount of work reiserfs does
>on each stat.
>
>atimes are used by many mail programs to find out the last time you read
>your mailbox file. nodiratime probably won't affect most applications.
>
>-chris
>
>
>
>
>
>
profiling will also tell you if the performance problem is in user
space.....;-)
--
Hans
next prev parent reply other threads:[~2003-12-01 7:47 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <bq8f7u$ufn$1@sea.gmane.org>
2003-11-28 16:46 ` Strange errors with 2.4.22 patches (from Chris) and bonnie++ Hans Reiser
2003-11-29 19:27 ` Christian Mayrhuber
2003-12-01 13:30 ` Chris Mason
2003-12-01 16:00 ` Christian Mayrhuber
2003-12-01 17:21 ` slow stat() ? David Bernick
2003-12-01 17:30 ` Chris Mason
2003-12-01 7:47 ` Hans Reiser [this message]
2003-12-11 4:44 ` update on " David Bernick
2003-12-01 17:43 ` David Bernick
2003-12-01 17:57 ` Chris Mason
2003-12-02 11:57 ` David Bernick
2003-12-01 8:02 ` Strange errors with 2.4.22 patches (from Chris) and bonnie++ Jens Benecke
2003-12-01 13:49 ` Chris Mason
2003-12-02 8:35 ` Jens Benecke
2003-12-03 19:13 ` Chris Mason
2003-12-04 12:44 ` Jens Benecke
2003-12-04 16:28 ` Chris Mason
2003-12-04 16:33 ` Chris Mason
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=3FCAF1F4.6040106@namesys.com \
--to=reiser@namesys.com \
--cc=bernz@bernztech.org \
--cc=mason@suse.com \
--cc=reiserfs-list@namesys.com \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.