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

Re: Improving the performance of libsvn_wc for checkouts/updates/switches

From: Josh Pieper <jjp_at_pobox.com>
Date: 2004-05-23 01:05:39 CEST

Josh Pieper wrote:
> Philip Martin wrote:
> > Josh Pieper <jjp@pobox.com> writes:
> >
> > > Is there a big reason to expect that if you 'kill -9' a checkout it
> > > will remember anything it did?
> >
> > It does now, so I would expect it to do so in the future.
>
> Do you see any correctness or data loss concerns with the approach
> other than this? If not, then I'll do some benchmarking and see if
> it is worth arguing this point. ;)

Results of the benchmarking are in. I did a checkout of a freshly
imported mozilla source tree as something of a real world test, plus
the same old synthetic things I had before.

The improvement on mozilla was 10% for BDB, 16% for FSFS.

         BDB FSFS
       ------------------
Trunk | 21m5s | 18m37s
       ------------------
O(n)WC | 18m49s | 15m34s
       ------------------

The synthetic ones are the same as I showed before, about a 4x speedup
once you get over 100 files in a directory.

Does anyone think this scale of performance improvement is worth
having the client have to re-download information if it is 'kill -9'd
or has a system crash? Note that it will not have to re-download
anything if CTRL-C or 'kill -15' is used.

If the vote is no, I'll shelve the patch and this performance
improvement will still be there for the taking by using the
entries+changes concept.

-Josh

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun May 23 01:06:06 2004

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