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

Re: Performance issues versus CVS

From: Brandon Ehle <azverkan_at_yahoo.com>
Date: 2004-04-16 09:28:03 CEST

Ben Collins-Sussman wrote:

>On Thu, 2004-04-15 at 11:32, Keven Ring wrote:
>>Has anyone seen performance issues such as these?
>Keven actually mailed me privately about this stuff, before I asked him
>to post this mail to the dev list. My comments were:
> * Hm, this looks like what we saw last summer, before we did the huge
>speed improvements in the fall. But after the improvements, checkouts
>were very, very close in speed to CVS. (Did anyone notice that since
>the 1.0 release, we've had no speed complaints? That's a big deal in my
> * I would expect svn imports to be slower than cvs, if only because
>the client *and* server are both doing compression. A fairer test might
>be to run 'cvs -z9'. But imports are generally one-time things, so the
>checkout/update timings are much important, I think.
> * Is anyone else seeing this sort of slowness? Might there be
>something specific about Keven's setup that's causing this?

I last ran my perf suite in Oct 2003 (I think right after a major
performance enhancement had been committed) and my results were similar.



The key thing I've gleamed from these tests is that you have Write Cache
Enabled (WCE) on your hard drives, preferrably use EXT3 or XFS for your
partitions (never got a chance to test Reiser4), hopefully use the 2.6
kernel, and have --bdb-txn-nosync enabled on your repositories to
achieve maximum performance.

I haven't formalized my results yet, but SVN running on an AMD64
architecture is also a nice performance boost. The only downside (and
reason for my results taking so long to finish) is that AMD64 driver
support is still very hard to come by.

To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Apr 16 09:28:27 2004

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.