[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Subversion on various Linux file systems performance test

From: Brandon Ehle <azverkan_at_yahoo.com>
Date: 2003-11-20 19:39:00 CET

Julian Foad wrote:

> Brandon Ehle wrote:
>
>> Here is the results of my Subversion benchmark running checkout with
>> the working copy and repository located on various file systems.
>>
>> The major points that I've gleamed from this test.
>>
>> * Linux 2.6 will provide a fast host for Subversion repositories when
>> its ready.
>> * JFS, FAT32, & ReiserFS make poor working copy hosts (client side).
>> * All of the file systems performed well for hosting the repository
>> (server side).
>> * The version of ReiserFS used in this test is 3.6. I would like to
>> repeat this test with Reiser4 when it starts merging into the kernel.
>>
>> http://subversion.kicks-ass.org/svn-fs-bench/
>
>
> Interesting. I'm using ReiserFS 3.6.9 on kernel 2.4.21 (SuSE 9); it's
> a shame you don't have that combination. I am surprised to see
> ReiserFS performing so badly for a WC, but I see that FAT32 is also
> much slower for a WC on kernel 2.6.test9 than it was on kernel 2.4
> (and yet much faster for the server), so maybe ResiserFS is not too
> bad on kernel 2.4.
>
> I don't know how far advanced the kernel 2.6 is. Is it possible that
> it will yet be "tweaked" or "fixed" so that the performance on FAT32
> WCs (and hopefully ReiserFS) is not so much worse than 2.4 was?
>
> Did you repeat these tests and check that there was not significant
> variation due to cacheing etc.?
>
I umounted and remounted all respective file systems between tests,
which I'm told (and seems to work), will effectively blow out away and
cached files.

I'm not really sure if any development effort was focused on the version
of ReiserFS that is in 2.6 or if most of the development effort went
into making Reiser4. Assuming the later is the case, we should see
massively different results with that test. It could be something as
simple as the default mkfs options that are in the 2.6 kernel haven't
been tweaked yet, or they have been tweaked to provide the best server
performance. ReiserFS takes second in the 2.6 repository test behind
EXT3 for the repository performance.

If you have a link to the kernel source that you use (hopefully not 400
hundreds patchs that I'll manually have to apply because I'm not running
Suse), I'll give it a whirl and see how the 2.4 ReiserFS performance
stacks up.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Nov 20 19:39:53 2003

This is an archived mail posted to the Subversion Dev mailing list.