| JMeter Memory Issue with Large Files |
|
| Saturday, 14 October 2006 | |
|
Just wanted to pass along a JMeter problem I encountered today so you don't spend a lot of time on it, especially since I've been promoting JMeter as a performance testing tool for PeopleSoft my last three posts. :-( Today I was trying to do some performance testing for our Enterprise Learning Management application -- basically testing to see if the wide area network was going to be able to handle video content that we'd be pushing once ELM goes live. I had a simple JMeter test script running on my workstation which called URL's to eight 26MB files. I was trying to test 25 concurrent user sessions which ramped up over 5 minutes. And that's where JMeter let me down. It seems that downloading large files takes more memory than small files. It bombed out with memory problems, so I increased the Heap size in the batch file that launches JMeter, and fired it up again. It ran the next time, but my memory went to zero and started paging. With my hard drive thrashing, performance on my workstation was understandably poor and my results were seriously skewed. After some research on the web and discovering it's actually a common problem, I went back to my old standby, Microsoft Web Application Stress tool. MS WAS doesn't do a good job with PS sessions in Tools 8.4x (as I recall), but since I was just testing eLearning content which is served outside of PeopleSoft, MS WAS worked fine. I had a script up and running in no time, it ran 25 users concurrently for 30 minutes with no memory problems. Since JMeter manages PeopleSoft sessions so well, I'll still use it for performance tests, but I'll plan to run it on multiple workstations with lots of RAM if I ever have to do serious performance testing. |
|
| Last Updated ( Saturday, 14 October 2006 ) |
| < Prev | Next > |
|---|