You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@tomcat.apache.org by Allistair Crossley <Al...@QAS.com> on 2004/12/07 15:28:52 UTC
RE: Performance with/without loadbalancing
One reason may be because in a load test you are creating *lots* of request threads in Tomcat. If you only have 1 Tomcat, you need to pay respect to the counters you have set on your Connector (maxThreads="150" minSpareThreads="25" maxSpareThreads="75"). If your load test exhausts these and all spares and the queue, you'll get connection denied. If you have 3 Tomcats and enough processing power on your server you are effectively multiplying up these request processors.
That's one theory anyway. Anyone?
> -----Original Message-----
> From: Shapira, Yoav [mailto:Yoav.Shapira@mpi.com]
> Sent: 07 December 2004 13:55
> To: Tomcat Users List
> Subject: RE: Performance with/without loadbalancing
>
>
>
> Hi,
> Did your testing properly account for ramp-up?
>
> Yoav Shapira http://www.yoavshapira.com
>
>
> >-----Original Message-----
> >From: Michael Südkamp [mailto:michael.suedkamp@docware.de]
> >Sent: Tuesday, December 07, 2004 7:50 AM
> >To: tomcat-user@jakarta.apache.org
> >Subject: Performance with/without loadbalancing
> >
> >Hi,
> >
> >I set up loadbalancing with Tomcat 4.1.30 and the JK
> connector and tested
> >it
> >with JMeter.
> >For a certain test configuration I found the following
> average request
> >durations:
> >
> >Tomcat workers request duration
> >-------------------------------------------
> >1 (no load balancing) 35 seconds
> >2 4 seconds
> >3 2,2 seconds
> >
> >All workers ran on just one PC. What I wonder is why a loadbalanced
> >configuration is that much more performant than a Tomcat
> configuration
> >without loadbalancing.
> >
> >Michael
> >
> >
> >---------------------------------------------------------------------
> >To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
> >For additional commands, e-mail: tomcat-user-help@jakarta.apache.org
>
>
>
>
> This e-mail, including any attachments, is a confidential
> business communication, and may contain information that is
> confidential, proprietary and/or privileged. This e-mail is
> intended only for the individual(s) to whom it is addressed,
> and may not be saved, copied, printed, disclosed or used by
> anyone else. If you are not the(an) intended recipient,
> please immediately delete this e-mail from your computer
> system and notify the sender. Thank you.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: tomcat-user-help@jakarta.apache.org
>
>
<FONT SIZE=1 FACE="VERDANA,ARIAL" COLOR=BLUE>
-------------------------------------------------------
QAS Ltd.
Developers of QuickAddress Software
<a href="http://www.qas.com">www.qas.com</a>
Registered in England: No 2582055
Registered in Australia: No 082 851 474
-------------------------------------------------------
</FONT>
---------------------------------------------------------------------
To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: tomcat-user-help@jakarta.apache.org
AW: Performance with/without loadbalancing
Posted by Michael Sudkamp <mi...@docware.de>.
>
> what I would do is to try a few things.
>
> 1. increase the ramp up time and see if that affects it. It could be
> the load balancer is helping tomcat handle the load
>
My ramp up is sufficient.
> 2. build the latest JMeter and use the distribution graph I wrote to
> look at the requests. the new graph will show the 50% and 90% line. If
> there's any pattern in the response times, you'll see it. What you
> should see is clumping of the data.
I run the test until I see a constant response time. Normally after about 10
minutes.
What other pattern should I see?
>
> 3. increase the number of threads to match exactly the number of
> threads you have in JMeter.
>
I have 50 threads. That should be lower than the maximum connector threads
(75). Should be OK, right?
After each request there is a pause of 10 s (constant timer). This sounds
like low load but the requests are quite processing intensive.
With only one worker, the CPU is not at maximum load. With three workers the
CPU goes up to 80/90 percent.
Michael
---------------------------------------------------------------------
To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: tomcat-user-help@jakarta.apache.org
Re: Performance with/without loadbalancing
Posted by Peter Lin <wo...@gmail.com>.
perhaps black magic?
what I would do is to try a few things.
1. increase the ramp up time and see if that affects it. It could be
the load balancer is helping tomcat handle the load
2. build the latest JMeter and use the distribution graph I wrote to
look at the requests. the new graph will show the 50% and 90% line. If
there's any pattern in the response times, you'll see it. What you
should see is clumping of the data.
3. increase the number of threads to match exactly the number of
threads you have in JMeter.
I hope that helps
peter
> > >Tomcat workers request duration
> > >-------------------------------------------
> > >1 (no load balancing) 35 seconds
> > >2 4 seconds
> > >3 2,2 seconds
> > >
---------------------------------------------------------------------
To unsubscribe, e-mail: tomcat-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: tomcat-user-help@jakarta.apache.org