public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
From: David Chinner <dgc@sgi.com>
To: Stephen So <s.so@griffith.edu.au>
Cc: xfs@oss.sgi.com
Subject: Re: Slow performance when extracting tarballs
Date: Tue, 1 May 2007 09:27:55 +1000	[thread overview]
Message-ID: <20070430232755.GT32602149@melbourne.sgi.com> (raw)
In-Reply-To: <4635DAA4.4070402@griffith.edu.au>

On Mon, Apr 30, 2007 at 10:01:40PM +1000, Stephen So wrote:
> Hi everyone,
> 
> Just a question about XFS when extracting bzipped tarballs containing
> lots of little files (e.g. the linux kernel source).  I've noticed on my
> new laptop (Intel Core 2 Duo @ 2.16 GHz), which has FC6 i386, which has
> XFS partitions that when I extract these types of tarballs, my system
> becomes rather non responsive, my mp3 player starts skipping etc.  After
> looking at top during the extraction process, bzip2 uses about 80-90% of
> CPU initially and extraction seems quite fast but after a few seconds,
> it drops to 30-40%, the system becomes non responsive, and extraction is
> much slower. 

The log probably fills up and then it falls back to the speed that
your disk can write back all the metadata.

> I created a new partition of ext3 and reiserfs, did the same tarball
> extraction, and on both filesystems, bzip2 uses at least 90%, extraction
> is fast the whole way through, and the system is quite responsive.
> 
> I created my XFS partition using the following command (I made a larger
> log file size since I heard that improve delete performance a bit):
> 
> mkfs.xfs  -l size=64m /dev/sda10
> 
> Then to mount this partition, I have these switches in my /etc/fstab file
> 
> noatime, nodiratime, logbufs=8

If you don't care about the filesystem always being able to recover
correctly when power fails (i.e. can lead to filesystem coruption
on power failure) you can also use the "nobarrier" option which
can significant;y speed up metadata perfromance on XFS.

> I'm using kernel 2.6.20 that came from the FC6 updates repositories.  So
> is there something wrong with my XFS setup?  Is my log file too small? 
> Or is this "normal" behaviour of XFS (i.e. that it excels best when
> working with very large files but not lots of little files)?

XFS excels at large files and/or lots and lots of files. On small files it
performs adequately but is not the fastest filesystem around.

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

      parent reply	other threads:[~2007-04-30 23:28 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-04-30 12:01 Slow performance when extracting tarballs Stephen So
2007-04-30 21:35 ` Chris Wedgwood
2007-05-04  2:45   ` Stephen So
2007-04-30 23:27 ` David Chinner [this message]

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=20070430232755.GT32602149@melbourne.sgi.com \
    --to=dgc@sgi.com \
    --cc=s.so@griffith.edu.au \
    --cc=xfs@oss.sgi.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox