[Benchmarking] Running the tests on the command line and summarizing results

[Andrea Aime]

Frank Warmerdam ha scritto:
> Andrea Aime wrote:
>> Once you have those you can run a test using:
>> jmeter -p jmeter.properties -n -t script.jmx -l script_results.jtl
>>
>> The jtl file is actually just a csv file with details of all requests,
>> you can run the summarizer on it to get a table summary with average 
>> time, throughput and so on, for example:
> 
> Andrea,
> 
> I'm getting an error and my .jtl file seems to be xml:
> 
> <?xml version="1.0" encoding="UTF-8"?>
> <testResults version="1.2">
> <httpSample t="274" lt="274" ts="1253126575692" s="true" lb="1" rc="200" 
> rm="OK" tn="1 1-1"
> dt="bin" by="5410">
>   <assertionResult>
>     <name>ContentTypeCheck</name>
>     <failure>false</failure>
>     <error>false</error>
>   </assertionResult>
> </httpSample>
> ...
> 
> Is there a jmeter switch to force csv output?  I'm using:
> 
>   jmeter -n -t bigtiff_fcgi.xml -l bigtiff_fcgi.jtl
> 
> for instance.

Hi,
if you look in /opt/scripts you'll see there is a checkout
of http://svn.osgeo.org/osgeo/foss4g/benchmarking/scripts
with the full set of tools to runa  script (besides java
and jmeter, that is).

I've made up the script that performs a full run + summaries
generation and tested it a bit, seems to be working fine.

There is a readme.txt file that show how to run it, but
it's really pretty simple:

./benchmark mapserver/ecw.xml

should run and give you a summary of results (provided you
add your own ecw.xml file there, that is).

As for GS performance, I get something between 9 and 10
images per second, but we've discovered a serious performance
regression in the mosaic code late in 1.7.x series (sigh) that
will be fixed in 1.7.7 and 2.0.x. There the performance should
be 30-50% higher, but I still have to check it out on the
benchmarking server.

Cheers
Andrea


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