[Benchmarking] 6600 requests csv and jmx files ready

[Andrea Aime]

Hi,
so during the meeting we discussed the fact that some servers
are disk bound, some others are not, which makes an order
of magnitude difference in the results.

One way out of that is probably to have the three runs use 6600
different requests: a single jmx file that internally makes
the three runs starting from 1-2-4-8-16-32-64, and then back
  1-2-4-8-16-32-64, and then again, using a single csv file
containing 6600 requests. That should make it generate only
unique requests.

Well... there is a catch that the reprojected tests use exactly
the same area (so that we can see the reprojection overhead)
but maybe with 6600 requests the OS won't be able to cache the
data.

Anyways, I've gave it a spin and tried to generate .jmx and
.csv files that follow the above design.

The are committed in svn, the jmx files have a name that
ends with 3times and the csv files with _3.
http://svn.osgeo.org/osgeo/foss4g/benchmarking/scripts/csv/
http://svn.osgeo.org/osgeo/foss4g/benchmarking/scripts/template/

I did not have time to test them out though.
Can someone do it?

And we also need to vote whether to make this change
last minute or not. I'm +0, yes, I think it's a good idea,
no, I feel it's late. But if everybody has a chance to rerun
the benchmarks before the presentation why not, at least
we won't have people wondering why there are 10 times differences
between the results, though we'll probably have people wonder
why the results are so similar (well, with the exception of
the ecw based tests in which of course Erdas will be the
clear winner not being disk bound with those ;-) )

Cheers
Andrea

-- 
Andrea Aime
OpenGeo - http://opengeo.org
Expert service straight from the developers.