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

Re: Subversion 1.9.0-dev FSFS performance tests

From: Stefan Fuhrmann <stefan.fuhrmann_at_wandisco.com>
Date: Mon, 30 Jun 2014 16:51:22 +0200

On Mon, Jun 30, 2014 at 4:06 PM, Ivan Zhakov <ivan_at_visualsvn.com> wrote:

> On 19 June 2014 14:21, Ivan Zhakov <ivan_at_visualsvn.com> wrote:
> > Hi,
> >
> > I've performed several FSFS performace tests using latest Subversion
> > from trunk_at_r1602928.
> >
> I've re-ran my FSFS performance tests with trunk_at_r1605444 using latest
> fsfs7 performance fixes including combining indexes to revision files.

Makes sense, that's close enough to what I used
as a basis last week. Optimizations since then have
mostly brought down the CPU overhead when reading
from SSD, RAM disk or disk cache.

> This time I used the same Windows VM hosted on Macbook with SSD disk.

Haven't looked at the results there, yet.

> Also I run the same tests on Windows VM on typical IBM x3620 M3 server
> with SAS raid attached spinning disks on Windows Hyper-V: this is very
> typical configuration in SMB environment.

Yes, that is certainly a typical configuration.

> Please find results in attached pdf for easier reading. I leave making
> conclusions to readers.

Could you re-run the dump tests with -r0:5000 as you did
two weeks ago. Otherwise, it will be hard to check the
results for plausibility (I don't have access to my server
hardware for the next 3 months or so).

Also, it seems that some of these tests are run from hot
caches - causing a lot of variation and making comparison
pointless. An extreme case:

ptime 1.0 for Win32, Freeware - http://www.pc-tools.net/
Copyright(C) 2002, Jem Berkes <jberkes_at_pc-tools.net>

=== "svn log http://localhost/svn/ruby-fsfs6-unpacked >nul" ===

Execution time: 216.064 s
Execution time: 13.268 s
Execution time: 18.061 s

I'm not pretending that my performance tests are scientific, but my
> goal is just to validate that fsfs7 is not worse than fsfs6.

That's a good goal. I'm currently through with my FSFS
TODO, except for documentation and committing a change
you suggested in Berlin. I'll then write a post on the pitfalls
of repo performance testing (plus a script to avoid them).
Furthermore, I'll convert the raw measurement data (plus
additional tests run later) into something easier to consume.

-- Stefan^2.
Received on 2014-06-30 16:51:54 CEST

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.