On 18/02/2010 21:58, kmradke_at_rockwellcollins.com wrote:
> Greg Stein<gstein_at_gmail.com> wrote on 02/18/2010 03:31:19 PM:
>> Awesome work! We've just been assuming/hoping it would be fast enough,
>> and would resolve any problems "later". It is good to see we're in the
>> right ballpark.
>>
>> On Thu, Feb 18, 2010 at 13:21, Philip Martin<philip.
>> martin_at_wandisco.com> wrote:
>>> How fast are sqlite queries be compared to the old-wc entry caching?
>>> Operations like update benefit from faster locking, but in old-wc
>>> status does no locking at all. Also the multiple entries files give
>>> the data a tree-like organisation that is not modelled quite so well
>>> by an SQL table. Is sqlite going to be at least as fast as old-wc at
>>> running status when the OS cache is hot? I've been vaguely assumming
>>> that it would be OK but this week I felt it was time to investigate.
>>> What would we do in 3 months time if wc-ng turns out to be an order of
>>> magnitude slower than old-wc?
>>> ...
>
> What platform were these test executed on? We need to make sure
> windows platforms are just as zippy.
>
> Kevin R.
For me on CYGWIN_NT-6.0-WOW64 brahe 1.7.1(0.218/5/3) 2009-12-07 11:48
i686 Cygwin
$ svn --version
svn, version 1.6.9 (r901367)
Create the test repo using the shell script, repeat "$ time svn status"
a few times:
real 0m37.303s
real 0m15.754s
real 0m15.832s
Create "wcx.db" using the python script, repeat "time
simplesqliteclient.exe" a few times:
real 0m0.107s
real 0m0.100s
real 0m0.093s
I'm not really in a position to patch svn 1.6 to stop it from doing the
actual file accesses, so I can't do the last test that Philip performed.
Great results!
Matthew
Received on 2010-02-19 18:45:27 CET