like wise I haven't seen this behavior. when I've run benchmarks against tomcat, I run tests for 8+ hours with over 1million requests.
you might want to try the distribution graph and see what is happening to get more information. peter On 4/18/06, Oliver Erlewein (DSLWN) <[EMAIL PROTECTED]> wrote: > > Hi @ all, > > I run a 3h perf test on production and after a constant amount of > samples (~5000 i.e. ~2h into the test). I get a huge bump in my return > times. This does not seem to be the application but jmeter/java getting > into a tangle. After about 12min the results normalise again. Now I've > run the same test in 3 parts of 1h directly back to back and I don't see > the bump. The Jmeter machine has more than enough grunt to handle the > test run. There are parameters in the jmeter and jmeter.bat that I'm > trying to configure. Tests this evening will show whether that will do > any change but maybe someone has played around with the parameters and > has some values that allow for long running big tests. > > The "bump" looks something like this: > > 60s + > + > 30s ++ + > ++ +++ > +++++++ > 6s ++ + ++ + + + + ++++++++ + + + + > 2s ++++++++++++++++++++++++++++++++++ > > Regards & Thank you > Oliver > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >