[prev in list] [next in list] [prev in thread] [next in thread] 

List:       subversion-dev
Subject:    Re: [svnbench] Revision: 1520942   compiled Sep  9 2013, 00:22:04 on x86_64-unknown-linux-gnu
From:       Neels Hofmeyr <neels () elego ! de>
Date:       2013-09-23 10:33:25
Message-ID: 20130923123325.4536270c () dub ! nat ! stsp ! name
[Download RAW message or body]


neels@apache.org:
> Charts of this data are available at
> http://svn-qavm.apache.org/charts/

Hi all,

I'm slightly confused by the current charts for 1.8. They suggest
that trunk was drastically slower than 1.8 the closer you get to the
branch point. That makes no sense, it should start almost even.=20

Also, the first four very negative runs in the new 1.8 charts were all
launched manually, at a very different time of day and day of week than
the cronjob usually runs them. It seems "things" are at work.

After this and the previous misinterpretation of a hardware upgrade I
didn't know about, I believe by now that it is somewhat futile to run
benchmarks on a virtualized host. The benchmarks seem to run at a
quiet time of the week, so we do get some thin bit of substance in the
data. But if even the time of day can jolt the numbers beyond
recognition, that suggests that the benchmarks are running in the wrong
place...

What do you think? Is this madness or is it ok as it is?

Thanks,
~Neels

["signature.asc" (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]