On Tue, Jun 14, 2011 at 5:56 PM, Mark Phippard <markphip_at_gmail.com> wrote:
> On Tue, Jun 14, 2011 at 5:48 PM, Johan Corveleyn <jcorvel_at_gmail.com> wrote:
>>
>> Also, has anyone tested this on an NFS-working copy? Or CIFS?
>
> AFAIK, no one is doing any performance testing on NFS or CIFS. I have
> repeatedly invited users to run the benchmarks I wrote in this
> configuration but no one has bothered. This leads me to believe it is
> not important to people.
I ran the benchmark tests on CIFS[1] with 1.6.17 and trunk_at_1116051 and
posted the results to https://ctf.open.collab.net/sf/projects/csvn
The good news is that 1.7, while obviously slower than a local drive,
is 40% faster overall compared to 1.6.17:
Basic Tests:
1.6.17 1.7_at_1116051
Action Time Time Diff
------------- ----------- ----------- ----------------
Checkout: 00:07:58.22 00:07:10.19 00:48.03 10.04%
Update: 00:07:55.75 00:04:23.66 03:32.10 44.58%
Switch: 00:00:20.97 00:00:03.99 00:16.97 80.95%
Proplist: 00:00:19.78 00:00:00.17 00:19.61 99.14%
Status: 00:00:14.15 00:00:01.70 00:12.45 87.98%
Commit: 00:00:52.26 00:00:16.47 00:35.79 68.48%
svnversion: 00:00:13.06 00:00:06.69 00:06.36 48.74%
Merge Tests:
1.6.17 1.7_at_1116051
Action Time Time Diff
------------- ----------- ----------- ----------------
Merge: 00:01:38.44 00:00:57.14 00:41.29 41.95%
Revert: 00:02:46.94 00:00:38.20 02:08.73 77.11%
Merge: 00:01:30.95 00:00:54.09 00:36.86 40.53%
Reintegrate: 00:04:36.04 00:01:38.31 02:57.73 64.39%
Folder Tests:
1.6.17 1.7_at_1116051
Action Time Time Diff
------------- ----------- ----------- ----------------
Checkout: 01:19:16.65 01:19:08.90 00:07.76 0.16%
Status: 00:03:00.93 00:00:53.52 02:07.41 70.42%
Commit: 00:04:32.14 00:03:33.72 00:58.42 21.47%
Update: 00:04:34.08 00:00:45.76 03:48.32 83.31%
svnversion: 00:03:25.66 00:01:38.84 01:46.81 51.94%
Switch: 00:06:59.56 00:01:11.68 05:47.88 82.92%
Commit: 00:04:45.71 00:03:00.06 01:45.66 36.98%
Merge: 00:07:54.19 00:01:10.17 06:44.03 85.20%
Reintegrate: 00:22:38.31 00:03:10.71 19:27.60 85.96%
Binaries Tests:
1.6.17 1.7_at_1116051
Action Time Time Diff
------------- ----------- ----------- ----------------
Checkout: 00:22:37.94 00:25:20.78 00:02:42.85 -11.99%
Update: 00:42:29.37 00:04:22.99 00:38:06.39 89.68%
Switch: 00:13:48.21 00:16:39.67 00:02:51.46 -20.70%
Cleanup: 00:00:41.17 00:00:40.56 00:00:00.61 1.48%
Delete: 00:00:58.97 00:00:09.91 00:00:49.06 83.20%
Status: 00:00:37.36 00:00:02.98 00:00:34.38 92.02%
Commit: 00:01:42.34 00:01:09.78 00:00:32.56 31.81%
Rename: 00:02:11.48 00:00:17.63 00:01:53.85 86.59%
Status: 00:00:35.52 00:00:02.64 00:00:32.89 92.58%
Commit: 00:02:09.61 00:01:10.84 00:00:58.77 45.34%
svnversion: 00:00:32.59 00:00:29.36 00:00:03.23 9.91%
Switch: 00:03:18.64 00:03:10.80 00:00:07.83 3.94%
Commit: 00:00:42.92 00:01:16.44 00:00:33.52 -78.12%
Merge: 00:12:17.98 00:01:29.62 00:10:48.35 87.86%
Reintegrate: 00:10:39.32 00:00:34.32 00:10:05.00 94.63%
TOTAL: 04:40:57.17 02:47:42.28 01:53:14.89 40.31%
The one glaring difference is the update test in the binaries tests;
throw that out and the difference is only 13%. I'm rerunning the
binary tests again to see if that variance repeats. Regardless, it
doesn't appear we are any worse off than with 1.6.17.
Paul
[1] Technically SMB 2.1 since both machines are Windows 7.
P.S. I gave a shot at testing over NFS. I tried to find a working
combination of cygwin, Allegro NFS Server, Omni-Lite, and/or Omni-NFS
Server (and probably a few others I'm forgetting) but couldn't get it
to work. I was hindered by the fact that Windows 7 only includes the
Client for NFS service in Win 7 Enterprise and Ultimate...and my
stable of machines tops out at Pro (and I wasn't too keen on doing an
in-place upgrade).
> I tried to do it myself, but the NFS performance in my environment was
> so slow that I did not have the patience to wait for even the 1.6
> version of the tests to finish.
>
> --
> Thanks
>
> Mark Phippard
> http://markphip.blogspot.com/
>
Received on 2011-06-16 17:57:51 CEST