1/ Try different java arguments. The default one "-Xincgc" is pretty poor. Something like this will probably improve things:
-Xms256m -Xmx256m -XX:NewSize=64m -XX:MaxNewSize=64m -XX:MaxLiveObjectEvacuationRatio=40 -XX:SurvivorRatio=8


2/ Avoid having your samplers download the images. Downloading the images requires parsing the HTML code, which is an expensive task.

Hope this helps,

Jordi.

En/na Tuomo Pyh�l� ha escrit:
Hi,

I tried jmeter and for me it seems to be somewhat slow. I tried running it
on 1.3GHz PC (Running RH Linux 9.0, JDK 1.4.2) , and I was able to get
throughput ~100req/min. For me this seems to be quite small amount of
requests! The machine running jmeter seems to be running out of cpu (cpu
usage ~100%). The machine running the server had a plenty of idle cpu
power. I wonder, what might be the reason for the behavior?

I tried running jmeter on other machines also (different OS), and
observed similar behavior. The test plan included some https-requests,
cookie-manager and authentication manager. I checked the requests to
retrieve the images also and added some assertions. I think ~100req/s
excludes retrieved images but still, it is a low amount of requests as
the pages have maybe 5 images or so per page. What should I do to improve
the performance?


--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]




---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]



Reply via email to