You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Brandon Ehle <az...@yahoo.com> on 2003/10/14 19:13:36 UTC
Scalability & Performance charts
I re-ran my old scalablity & performance charting scripts on subversion with rev 7404 and here's
the results I got. These charts are pretty noisy and subversion is a lot faster than a year ago
when I last ran these charts so I probably need to increase the dataset sizes to look for current
scalability issues. Take these results with a grain of salt until I can get some verification
done on the results.
These were done on a AMD 1.5GHZ box running RedHat 9 Kernel 2.6-test7.
The time charts are here:
http://subversion.kicks-ass.org/perf/elapsed.html
The full results including memory footprint (RSS) are here:
http://subversion.kicks-ass.org/perf/
I'm going to re-visit my perf analysis over the next two weeks to see where subversion is at
nowadays. I plan on fixing my ra_dav scripts and adding a couple comparisons to these charts. If
there is anything you'd like to see me add to my comparisons, let me know.
I'm currently interested in investigating:
* scalability charts of property queries, gets, and sets
* changing the apples vs oranges Perforce comparisons to apples vs apples comparisons
* Linux24 vs Linux26 vs Win2K on the same machine
* ra_dav with mpm-worker vs mpm-prefork
* hard drive write cache enabled vs disabled
* repository hosted on Ext2 vs Ext3 vs Reiser4 vs JFS vs XFS vs FAT32
* WC hosted on Ext2 vs Ext3 vs Reiser4 vs JFS vs XFS vs FAT32
__________________________________
Do you Yahoo!?
The New Yahoo! Shopping - with improved product search
http://shopping.yahoo.com
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Scalability & Performance charts
Posted by Philip Martin <ph...@codematters.co.uk>.
Brandon Ehle <az...@yahoo.com> writes:
> http://subversion.kicks-ass.org/perf/elapsed.html
It's interesting that NOSYNC has a huge effect on import, but does not
really have a significant effect anywhere else.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: Scalability & Performance charts
Posted by Sigfred HÃ¥versen <su...@mumak.com>.
On Tuesday 14 October 2003 21:13, Brandon Ehle wrote:
> I re-ran my old scalablity & performance charting scripts on subversion
> with rev 7404 and here's the results I got. These charts are pretty noisy
> and subversion is a lot faster than a year ago when I last ran these charts
> so I probably need to increase the dataset sizes to look for current
> scalability issues. Take these results with a grain of salt until I can
> get some verification done on the results.
>
Interesting results. I'm probably not alone in looking forward to the test
results for httpd. You did not specify whether you tested for the case of
access over LAN, but such results would also be welcome :-)
/Sigfred
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Scalability & Performance charts
Posted by ryan <ry...@netidea.com>.
Hi,
If I may offer some advice here...
If we take a use-case approach, I think we end up with an apples vs
apples comparison.
Here are the operations that me as a user of a large p4 repository find
important:
- checkout fresh tree (10,000 file checkout not entirely uncommon,
1000-5000 common place and to be expected)
- resync local tree with server (after my co-developers have checked in
many files)
- check in 100-300 files at a time (less common: check in up to 5000
files)
To me as a heavy user of source control, if these operations were
instantaneous then my developer time would be very happy indeed.
The charts suggest that SVN is actually a worse performer than CVS at
tasks like and similar to these. Hopefully it's from a really old
build :-)
Regards,
-ryan
On Tuesday, October 14, 2003, at 12:13 PM, Brandon Ehle wrote:
>
> I'm currently interested in investigating:
> * scalability charts of property queries, gets, and sets
> * changing the apples vs oranges Perforce comparisons to apples vs
> apples comparisons
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Re: Scalability & Performance charts
Posted by ryan <ry...@netidea.com>.
Hi,
If I may offer some advice here...
If we take a use-case approach, I think we end up with an apples vs
apples comparison.
Here are the operations that me as a user of a large p4 repository find
important:
- checkout fresh tree (10,000 file checkout not entirely uncommon,
1000-5000 common place and to be expected)
- resync local tree with server (after my co-developers have checked in
many files)
- check in 100-300 files at a time (less common: check in up to 5000
files)
To me as a heavy user of source control, if these operations were
instantaneous then my developer time would be very happy indeed.
The charts suggest that SVN is actually a worse performer than CVS at
tasks like and similar to these. Hopefully it's from a really old
build :-)
Regards,
-ryan
On Tuesday, October 14, 2003, at 12:13 PM, Brandon Ehle wrote:
>
> I'm currently interested in investigating:
> * scalability charts of property queries, gets, and sets
> * changing the apples vs oranges Perforce comparisons to apples vs
> apples comparisons
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org