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

Subversion Perforce with of BDB_TXN_NOSYNC enabled vs disabled

From: Brandon Ehle <azverkan_at_yahoo.com>
Date: 2003-02-27 16:08:03 CET

I've found some very interesting results in running a my performance
tests with BDB_TXN_NOSYNC.

The full list of results is here:
http://www.pipedreaminteractive.com/svn/

First off, BDB_TXN_NOSYNC, doesn't appear to affect anything other than
imports & commits.

While the times didn't change significantly with ra_dav, I see a huge
improvement for ra_local and ra_svn. In fact, if you look at this
chart, you can see that import performance of ra_local & ra_svn with
BDB_TXN_NOSYNC is nearly identical to Perforce's import performance!!!
http://www.pipedreaminteractive.com/svn/athlon/elapsed.html

I realize that its supposedly not safe to run with BDB_TXN_NOSYNC, but I
am going to conduct some real world multi-user testing with this flag
enabled and see what kind of results I get.

Looking at the import times with cachegrind, I now see that vdelta
accounts for 40% of the total execution time.
http://www.pipedreaminteractive.com/svn/import_nosync0.png

Here's the same view, but with the names expanded so you can read them
easier (although the area isn't correct here).
http://www.pipedreaminteractive.com/svn/import_nosync1.png

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Feb 27 16:07:07 2003

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.