jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Lin <wool...@gmail.com>
Subject Re: Performance problems
Date Wed, 07 Dec 2005 15:19:33 GMT
what is your target performance? or is the goal more general like find out
the max TPS. What kind of server hardware are you using?

Here are some general thoughts.

2.5-3Ghz system can handle roughly 25 concurrent parser processes.  the
performance depends on the size of the XML, but generally speaking, running
JMeter with more than 25 threads will be limited by XML performance. I
recommend using client systems that are atleast 2ghz and keep the thread
count under 20.

If the target server has multiple CPU's of the same speed or faster, you
will need atleast 1 client system per CPU on the server. Say I have a 4 CPU
server with 3.2Ghz.  To max out the server, I would need 6
2.4Ghzworkstations running JMeter.

hope that helps

peter


On 12/7/05, Ross Garrett <ross.garrett@mobilecohesion.com> wrote:
>
> Hi all
>
> Thanks for all the input.
>
> Unfortunately we're struggling to hit the performance figures we expect.
> We
> have tried deploying JMeter on a higher spec box, using multiple injectors
> and we're going to see if running in console mode will help significantly.
>
> If anyone has anything further to add it'd be much appreciated.
>
> Thanks again
>
> /Ross
>
> -----Original Message-----
> From: Benjamin Francisoud [mailto:benjamin.francisoud@anyware-tech.com]
> Sent: 02 December 2005 15:21
> To: JMeter Users List
> Subject: Re: Performance problems
>
> Using multiple test machines to test load on the server machine can help
> you too to achieve higher throughput... (but I never did it with jmeter)
>
> see: http://jakarta.apache.org/jmeter/usermanual/remote-test.html
>
> --
> Benjamin Francisoud
>
> Peter Lin wrote:
>
> >unfortunately, XML is heavy weight and the default xml parser shipped
> with
> >sun jdk is crimson. it's rather slow, so the practical limit per system
> is
> >roughly 20-40 depending on the CPU speed.  The only way to simulate
> larger
> >loads is to increase the think time, or use several client systems to
> >generate load.
> >
> >on a 1.4Ghz Pentium M laptop, the limit for soap performance is around
> >20-25. If you check use memory cache in the webservice soap sampler, you
> get
> >up to 50, but at that point the test is measuring XML performance on the
> >server.  hope that helps
> >
> >peter
> >
> >
> >On 12/2/05, Ross Garrett <ross.garrett@mobilecohesion.com> wrote:
> >
> >
> >>Hi
> >>
> >>I have just started using JMeter for performance testing of web service
> >>applications. I am currently using one thread group, and have been
> >>trying various numbers of threads. The maximum performance I can get
> >>(using 30-35 threads) is approx. 140tps which I know is just over half
> >>the current benchmark for the service under test.
> >>
> >>Should I be configuring JMeter differently to achieve higher
> >>performance, or will it just not go any faster?
> >>
> >>Thanks in advance
> >>
> >>/Ross
> >>
> >>---------------------------------------------------------------------
> >>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
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message