jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Stover <mstov...@apache.org>
Subject RE: JMeter 1.9.1 bottlenecked
Date Tue, 22 Jun 2004 20:57:49 GMT
I recommend you use the -server option.  I recommend NOT using -Xincgc.
ie

java -server -mx1538m ...etc

I don't know what listeners you are using, but I suggest that when you
feel ready to ramp up your test, remove all listeners except on
Aggregate Report listener.  It will use the least memory and cpu.

I would also try running two instances of JMeter and see if your cpu
usage stays at 51% and no appreciable gain in throughput.  If that is
the case, then the problem is outside of JMeter.

-Mike

On Tue, 2004-06-22 at 16:44, Liu, Julia wrote:
> First of all, thank you very much. 
> 
> In my case, there is a lot of room in terms of network i/o, CPU and memory.
> I monitored all those resource consumption while running the test. 
> 
> I need to add load up to 750 threads to reach throughput of 5400 order/hour
> against our application in order to meet performance requirement. Although
> running parallel instance of no-gui mode might help, it would be better to
> have one test result report without combining serveral.
> 
> Thanks a lot!
> Julia
> 
> -----Original Message-----
> From: Peter Lin [mailto:woolfel@gmail.com]
> Sent: Tuesday, June 22, 2004 2:13 PM
> To: JMeter Users List
> Subject: Re: JMeter 1.9.1 bottlenecked
> 
> 
> >From first hand experience, running a test with 50 threads is usually
> enough to saturate the network IO.  This is with static HTML and
> tomcat4/5.
> 
> one way to overcome this is to access the webserver from two different
> ethernet ports. For example, my dev environment at home include 4
> servers. My X1 and Linux box both have 2 ethernet cards.
> 
> one quick way to test if the network IO is the bottleneck is to use
> Apache AB to stress the server. If you get similar throughput with AB,
> then you know it's really network IO.
> 
> in general, the more threads you add beyond the max throughput will
> start to overload the system and reduce the throughput. What I like to
> do is always run a simple through put test against a given webserver
> to establish the ceiling. This way, you know what the absolute maximum
> is for a dynamic web application, since the static page will be the
> actual max.
> 
> I hope that helps
> 
> peter
> 
> 
> 
> On Tue, 22 Jun 2004 13:38:25 -0400, Liu, Julia
> <julia.liu@sigma-systems.com> wrote:
> > 
> > I start up JMeter with this option:
> > 
> > java -client -Xincgc -Xms1538m -Xmx1538m -jar `dirname
> $0`/ApacheJMeter.jar
> > "$@" on Solaris 8 machine, which has 2048m memory and 2x900 CPUs.
> > 
> > I start testing with 75 threads, CPU usage reaches 51%, then I increase
> > threads to 150, CPU usage remains 51%, no matter how many threads
> increased,
> > the highest CPU Usage remains the same: 51%. It looks like JMeter has
> > bottlenecked  some where. And also, the more threads you add, the lower
> the
> > throughput you get. Can someone explain how to fix this problem?
> > 
> > Thanks!
> > Julia
> > 
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> > 
> >
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
-- 
Michael Stover <mstover1@apache.org>
Apache Software Foundation


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


Mime
View raw message