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

Re: [svnbench] Revision: 1463069 compiled Apr 1 2013, 00:22:00 on x86_64-unknown-linux-gnu

From: Neels Hofmeyr <neels_at_elego.de>
Date: Tue, 2 Apr 2013 01:29:09 +0200

On Mon, 01 Apr 2013 00:44:12 +0000 neels_at_apache.org wrote:
> 1.7.0_at_1181106 vs. trunk_at_1462907
> Started at Mon Apr 1 00:26:11 UTC 2013
>
> *DISCLAIMER* - This tests only file://-URL access on a GNU/Linux VM.
> This is intended to measure changes in performance of the local
> working copy layer, *only*. These results are *not* generally true
> for everyone.
>
> Charts of this data are available at
> http://svn-qavm.apache.org/charts/

It's time for another chart reading summary, as there has been a
dramatic change that is shown to be consistent for the past three
benchmark runs.

Whoever is reading this, keep in mind that all of these findings only
apply to a local non-network connection to the repository, using
file:// URLs only, and even then: Your Mileage May Vary. Thank you.

Somewhere between r1454789 and r1457253 on trunk, that's three to
four weeks ago, 'commit', 'copy' and 'merge' got dramatically faster: as
much as 80%, and more! Amazing.

Commit and copy got faster across the board, merge only in the 5x5
working copy scenario (5x5 means: 5 dir levels, with 5 files and 5 more
subdirs per dir -- an evenly spread, "realistic" working copy).

The most noticeable speed increase is seen for a commit in the 5x5
working copy: of 1.7.0's average 25.5 seconds of runtime, trunk has
lost around 21 seconds! ... That's twenty-one full seconds!

A merge in a 5x5 WC with trunk has now lost about seven of 1.7.0's ten
seconds of runtime. 70%!

Four weeks ago, trunk's copy in a 5x5 WC has actually been slightly
slower than 1.7.0. Now it's 90% faster than 1.7.0, dropping from a
quarter of a second in 1.7.0 to near instantaneous completion on trunk.

The benchmark's grand totals now suggest that trunk's libsvn_wc is --
drumrolls -- on average 62% faster than 1.7.0! Just four weeks ago that
was "a mere" 25%.

Above speed gains have been consistent for the past three benchmark
runs, i.e. the past three weeks.

BTW, if you noticed some red bars for the same three test runs with a
speed *loss* of up to 90%, relax: all of those account for far less
than 0.1 seconds; just take a look on the right-hand-side charts.

I hope you enjoyed this summary from your friendly neighborhood chart
interpreter. Good work lads!!1!1 (It was definitely not me, that's for
sure.)

Meep meep!
  -- Roadrunner

~Neels

Received on 2013-04-02 01:29:55 CEST

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