Re: Comparing response times on local/remote tetsing

[Mike Stover <ms...@apache.org>]

I saw your first post - I honestly don't know the answer.  My only guess is the extra 
time comes from the serialization of all the SampleResults to be passed via RMI to 
the controlling machine.  50 threads are probably producing a lot of results that 
have to be serialized.  You could turn on DEBUG for log_level.jmeter.engine (in 
/bin/jmeter.properties) and see what the log says.  You'll probably see a lot of lines 
having to do with the result queue.  If it is consistently getting up to 200, then that 
may be the problem.

Are the times stable, or do they fluctuate?

-Mike

On 28 Jan 2003 at 22:27, Keld Helbig Hansen wrote:

> I've posted this before, but got no responses, so I'll try once more under a new subject :-)
> 
> I've done some experimentation with remote testing. It works fine, but I get some results that I don't fully understand. Maybe you do?
> 
> Here's the setup:
> 3 identical Win2000 computers, 512Mb RAM, approx. 500MHz.
> JDK is 1.4 on all machines.
> Number one running a Tomcat 4.1.12 with a simple servlet application (2 screens, 7 user functions). 
> Number two running the JMeter Server.
> Number three running the JMeter Client in GUI mode.
> Nothing else is running on the three machines.
> I've recorded a use case with 10 user interactions.
> 
> If I run non-remote mode (from computer three to computer one) with a single thread  I get response times around 55 msec. The JMeter computer (computer three) has 100% cpu.
> 
> Using remote testing I get respnse times around 40 msec. I can see that the JMeter server has 95% cpu, and the web server only 10% cpu. 
> I realize that it would be better if the CPU was not so busy on the JMeter server. But I get better response times with the remote testing setup, and this was what I expected.
> 
> Here comes the part that I can not explan.
> 
> Now I set 50 threads, and a JMeter timer delay between 1 and 3 seconds (to simulate a realistic load). 
> 
> On the non-remote setup I get response times around 200 msec. Cpu is 100% on the JMeter client (computer three) and 10% on the web server (computer one).
> 
> On the remote setup I get response times around 500 msec, 100% cpu on the JMeter server (computer two), 10% on the web server (computer one).
> 
> Why are the response times so much poorer with remote testing???????
> 
> It also seems that I need a real big machine for the JMeter server in order to put real load on the web server.   
> 
> 
> --
> To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
> For additional commands, e-mail: <ma...@jakarta.apache.org>
> 



--
Michael Stover
mstover1@apache.org
Yahoo IM: mstover_ya
ICQ: 152975688
AIM: mstover777

--
To unsubscribe, e-mail:   <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>