Hi all, I've got quite a big JMeter harness and it's still running Fedora 5 for the JMeter load drivers. Now I want to update those boxes to Ubuntu 8.04. i.e. I have upgraded one of the boxes and am running two against each other to see that nothing has changed. Well.... they are different. It looks like I have more variance in my results and the average response time is 50-100ms higher (about 15% of total average response time) with the new Ubuntu box with the same test plan.
I've checked most of the network settings and they are the same (if not better on Ubuntu). The boxes are physically the same and are on the same switch and network infrastructure. The only other difference is Java. Fedora runs 1.5.0_b07 and Ubuntu is 1.6.0_07. In my testing-Java oblivion I didn't pay too much attention to this previously but with differences what they are it becomes vital to me to track this deviation down. I have now installed Java 1.5.0_16 on Ubuntu and am running the tests again (will take a day though). If software/config can have such an impact on my perf test results I'm getting very concerned. This means I have to really watch out for exact installations and hardware. Not really something I want to go into... Anyone notice any differences with different Java versions? Anything else I should check? Or any other wild and wondrous ideas?! Cheers & Thanks! Oliver --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
