Will we be still including daytrader in the Geronimo assemblies?
Do you have any thoughts on using an open source tool to drive the load
testing that use a standard set of test plans/scripts. This would allow
anyone to easily run the benchmarks and ensure we are using the same
test plans (e.g. number of requests, the actual request sent) from
release to release.
AFAIK, ActiveMQ uses Apache JMeter for their JMS benchmarking.
http://svn.apache.org/viewcvs.cgi/incubator/activemq/trunk/jmeter/
http://activemq.org/Benchmark+Tests
http://activemq.org/JMeter+Performance+Tests
There may be some ideas we could take from there or the ActiveMQ guys
might have some ideas based on their JMeter experience.
"Long term" (maybe a few generations away..), we may want to look at how
we could automatically detect performance regressions.. E.G. GBuild
running benchmarks and comparing results with previous results in a
database.
John
Matt Hogstrom wrote:
All,
I have copied DayTrader from trunk and been restructuring it to stand
alone so it can be versioned independently from Geronimo. This way
performance comparisons can be made. It is currently located at
https://svn.apache.org/repos/asf/geronimo/daytrader.
You'll currently see a trunk dir in the daytrader tree. I am planning
in versioning it as 1.1 and removing it from trunk before we ship
1.1. So that means that I currently have the version set to
1.1-SNAPSHOT.
As far as the group is concerned from a Maven perspective I was going
to specify it as org.apache.geronimo.samples and not
org.apache.geronimo.samples.daytrader as I think it makes more sense
to leave the samples as independent and allow us to group them
independently if need be. I'm not sure if this is really right but
makes sense to me.
Once its complete we'll branch and release DayTrader 1.1 at the same
time as Geronimo 1.1.
So, once this is complete I want to start the discussion about
morphing DayTrader into an EJB 3.0 benchmark. So put your thinking
caps on about what would be a way to morph the benchmark into the next
generation. Some themes I'd like to capture is:
Easier to use
Better Documentation
EJB 3.0 and possibly 2.1 primitives
Round out a new set of JMX primitives
Improve the JMS primitives
Automatic build for multiple databases
Dynamically detecting cardinality and scaling appropriately
To start anyway.
Cheers,
Matt
