You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by Srinath vaidyanathan <sr...@gmail.com> on 2005/09/08 14:03:34 UTC

Simulating more than 1000 Threads from Jmeter

Hello,

I'm stress testing a web application using JMeter. I have 3 client machines 
at my disposal. I have read the new manual for distributed testing and i've 
manged to get it working with one of the machines as master and the other 2 
as slave. My question is this. "Suppose i want to simulate 3000 threads 
simulataneously can I do it ? Say, 1000 via each machine." 

Also I don't have many listeners in the test plan. i just have the simple 
data writer and view results as tree listeners.

My worry is that the latest distributed testing pdf manual says that 
simulating more than 300-500 threads is not advisable.

Therefore, How should I go about simulating that many threads? 

Thanks
Srinath Vaidyanathan

Re: Simulating more than 1000 Threads from Jmeter

Posted by Sergei Riaguzov <ri...@gmail.com>.
On 9/8/05, Srinath vaidyanathan <sr...@gmail.com> wrote:

Hello,
> 
> I'm stress testing a web application using JMeter. I have 3 client 
> machines
> at my disposal. I have read the new manual for distributed testing and 
> i've
> manged to get it working with one of the machines as master and the other 
> 2
> as slave. My question is this. "Suppose i want to simulate 3000 threads
> simulataneously can I do it ? Say, 1000 via each machine."


Well if you use Linux with at least 2.6.8 kernel and correct version of 
glibcs (Debian Sarge has the right one, Slackware has the wrong version as 
far as I heard) and JDK 1.5 and if you will have some 1-2 Gb of RAM on your 
computers it is possible without problems.

If you want to go further you should take care about not using BeanShell in 
your test plan, about trying to make your thread group not to complicated 
and so on.

And change Xms Xss values in jmeter startup script.

1500 threads is not a problem.

It is possible to go further with some time spent on configuration and so 
on.

Re: Simulating more than 1000 Threads from Jmeter

Posted by sebb <se...@gmail.com>.
On 08/09/05, Sergei Riaguzov <ri...@gmail.com> wrote:
> 
> 
> On 9/8/05, sebb <se...@gmail.com> wrote:
> 
> > Don't use remote mode - the client JMeter will have to handle all the
> > samples, which will stress it and perhaps the network. 
> > 
> > Use non-GUI (batch) mode on each of the machines, and combine the test
> > results after the run.
> Is there a JMeter native util to merge test results? How it handler same
> user names? I have to write a script to merge files but maybe there exists
> one?
> 

There is no utility, but it's not difficult to merge files.

For CSV files just use the appropriate OS "append" utility - not sure
if need to sort the file afterwards.

For XML, there is a header which would need to be removed from all but
the first file before appending them.

The combined CSV or XML file can then be read into JMeter in the usual
way, or processed by whatever you normally use to process the output.

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org


Re: Simulating more than 1000 Threads from Jmeter

Posted by Sergei Riaguzov <ri...@gmail.com>.
On 9/8/05, sebb <se...@gmail.com> wrote:

Don't use remote mode - the client JMeter will have to handle all the
> samples, which will stress it and perhaps the network.
> 
> Use non-GUI (batch) mode on each of the machines, and combine the test
> results after the run.

Is there a JMeter native util to merge test results? How it handler same 
user names? I have to write a script to merge files but maybe there exists 
one?

Re: Simulating more than 1000 Threads from Jmeter

Posted by sebb <se...@gmail.com>.
Don't use remote mode - the client JMeter will have to handle all the
samples, which will stress it and perhaps the network.

Use non-GUI (batch) mode on each of the machines, and combine the test
results after the run.

This has been discussed a few times on this mailing list.

Are you really sure you need 3000 threads?

S.
On 08/09/05, Srinath vaidyanathan <sr...@gmail.com> wrote:
> Hello,
> 
> I'm stress testing a web application using JMeter. I have 3 client machines
> at my disposal. I have read the new manual for distributed testing and i've
> manged to get it working with one of the machines as master and the other 2
> as slave. My question is this. "Suppose i want to simulate 3000 threads
> simulataneously can I do it ? Say, 1000 via each machine."
> 
> Also I don't have many listeners in the test plan. i just have the simple
> data writer and view results as tree listeners.
> 
> My worry is that the latest distributed testing pdf manual says that
> simulating more than 300-500 threads is not advisable.
> 
> Therefore, How should I go about simulating that many threads?
> 
> Thanks
> Srinath Vaidyanathan
> 
>

---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org