Re: Performance of ext4

[Holger Kiehl]

On Thu, 19 Jun 2008, Theodore Tso wrote:

> On Wed, Jun 18, 2008 at 05:58:00AM +0000, Holger Kiehl wrote:
> > For afdbench: 5336.41 files per second 15.63 MiB/s
> >
> > So it seems that for afdbench the ext4-patch-queue is a slowdown.
>
> Can you remind me where afdbench can be downloaded?  And if I remember
> correctly, it creates and deletes large numbers of small files,
> correct?
Yes and yes.

You can download the benchmark but it is complicated to setup. You can
download it from ftp://ftp.dwd.de/pub/afd/afdbench-2.0.0.tar.bz2 and
you will also need ftp://ftp.dwd.de/pub/afd/development/afd-1.4.0pre1.tar.bz2

Here a short guide how you need to set this up (there is also a SETUP
file in afdbench-2.0.0.tar.bz2):

   - create a new user for example afdbench
   - untar afdbench-2.0.0.tar.bz2 where ever your test filesystem is mounted
     eg /home
   - ln -s /home/afdbench-2.0.0 /home/afdbench
   - ensure that in .bash_profile of user afdbench you have:
        PATH=$PATH:$HOME/bin
   - login as afdbench
   - Untar afd-1.4.0pre1.tar.bz2
   - cd afd-1.4.0pre1
   - ./configure --prefix=$HOME --enable-ftp_reuse_data_port --enable-passwd_in_msg --enable-expand_path_in_message --enable-compiler-optimizations --enable-with_afdbench_settings --enable-splice_support --enable-sendfile_support
   - make
   - make install-strip
   - In afdbench script change BENCH_PASSWD to whatever you have set the
     password of user afdbench.

If you have problems because you do not have openmotif or lesstif, just use
the configure switch --with-gui=none. Also make sure you have an FTP-server
running, I always used vsftpd. To run the test I just called tiny-bench,
in it you will find how you can start it. You can also run without FTP-server
but I do not know if the problems are reproduceable.

> It would be interesting to see which new feature introduced by the
> ext4 patch queue --- probably dellayed allocation or mballoc --- is
> responsible for the slowdown.  One or the other (or both) can be
> disabled by mounting the filesystem (using a kernel with the ext4
> patch queue) with the mount options -O nomballoc or -O nodelalloc.
>
> If it turns out that nomballoc restores the speed for afdbench, for
> example, then it will tell us where we need to look more closely.
> Ideally we would not want to have one mount option needed to optimize
> filesystem operations for large amoutns of modifications to small
> files, and another mode of operation when mostly writing to large
> files.  So if you could do a round of tests using the ext4 patch queue
> kernel, with -O nomballoc and -O nodelalloc (and if both seem to
> improve things, try "-O nomballoc,nodelalloc" and see if you get back
> to the pre-ext4 patch queue speed), it would be very much appreciated.
Yes, I will try and redo the test as suggested, it might just take a while
since I just made my testing system operational.

What worries me more is the truncation of files, it makes the filesystem
unusable since you loose data. I hope there will be a solution for this.

Holger

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/