It might be useful to look at the min/max response times in the two cases.
The cheapest way to do this is to add a Summariser Post-Processor - this will generate summaries to the log every 3 minutes. The summary includes the min/max and average times. As to why you are seeing the difference - could there be some kind of caching occurring? How long do you run the test for? How many threads in each Thread group? How many loops? S. On 06/05/06, Tim Wood <[EMAIL PROTECTED]> wrote:
Hi, I am getting inconsistent results when I run a test multiple times without restarting jmeter between trials. I'm trying to see the maximum throughput I can get from a web server, so my test layout is simply two thread groups which repeatedly access a page on the server. The first time I run this test I get about 550req/sec throughput. If I repeat the test without exiting jmeter (ie. I just clear the results and hi= t run again) I get a rate of almost 1000 req/sec. From then on, I will consistently get about 1000req/sec. This is "consistently inconsistent" in that if I exit jmeter and restart it, I will get the same pattern - one ver= y poor run followed by almost 2x increase in performance. I also find that i= f I run the test using the non-gui mode of jmeter I will consistently get the low rate. Does anyone have any ideas of why this might be occuring? I'm only barely maxing out the server I am benchmarking when I get 1000 req/sec, so it does not seem like a limitation of the server in any way. My client and server are both linux 2.4ghz with 1gb ram, using java 1.5 and jmeter 2.1.1. Any ideas? Thanks! -Tim
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
