Re: High CPU utilization on Websphere Application Server

[Roland Jefferson JR]

Yes, test after hours when the other apps are sleeping.  If that doesn't work,  get management approval to shut off the other apps for your test Window.
-Roland

On Dec 13, 2012 10:01 PM, "rahul sinha" <rahulsi...@gmail.com> wrote:

Hello All,
  I had a question for all of you. I had to run a load test for a web based application and have set up a monitor using sitescope to monitor the CPU Utilization, memory and Unix Resources on the server. The issue i am facing is that the Websphere server  is shared by many apps and therefore even before beginning the load test the CPU utilization on the server would fluctuate between 50-80%. As expected, when i ran the load test i saw that the utilization was somewhere in the range of 86-92% throughout the test.

I advised the developer to increase the JVM memory size but this did not do the trick.
FYI: The counters for memory was 1% during the entire test and the Unix resources were also in the normal range.

Has anyone encountered a similar problem and can advise what might be the possible reason and solution to this problem.

Appreciate your help.

Thanks,
Rahul

--
 
 

[Manik Sikka]

Try to shutdown other tests/Have nothing running on the App server and there's your Answer...

Sent from my iPhone

[Mahesh Kumar Somaraju]

Is the production environment is also configured in the same way as test environment like sharing same Websphere for many apps??

If not then it is advised to shutdown the rest of apps and rerun the tests. It is always good to replicate the production.


Cheers,
Mahesh

--
 
 

[V.M.Guruprasath]

If I understand correctly, When CPU utilization is high and memory utilization is 1% why was memory adjusted?

-V.M.G

[Ruslan Kholyavkin]

    You cannot just adjust heap size to get better performance ( it may probably get it better ) but proper Heap Size , Perm Size  ,  Garbage Collection and compiler of JVM -- need to be tune / monitor and test in order to set proper setting and make sure that JVM by itself  not become a bottleneck.  Incorrect , not default Garbage Collection type,  can generate OOM by it self
 
Thanks,
Ruslan

Thanks for the reply Everyone.
On Friday, December 28, 2012 2:56:59 PM UTC-5, jayakri...@gmail.com wrote:

Hi Rahul,

After reading the above replies , I see solution is already came. But wanted to share my views on this.

1) Generally if you want to assess utiligation of your app server (web/DB) , make sure what ever the application you are going to test that should be up and running on WAS (since you are using WAS server), before you start your load test make sure no other app is running (if there is any compulsary call you should share the app server , than ask them for time slot and ask WAS admin to stop all apps which are ruuning on it except your app).

2) Also you mention JVM settings , these settings are related to memory usage not for CPU , we can change the JVM settings only when your app needs more memory than initally specified [ assume new settings is also filled up means means of 100% memoru usage , there could a memory leak , should be monitor JVM how it is behaving during your load test]

Thanks
Jaya K


\\+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

On Friday, December 28, 2012 2:09:08 PM UTC+5:30, Suresh wrote:

Hi Rahul,

R u doing load test in a) Testing environment or B) Production environment

A) Testing environment : stop all the App Server and do the load test

B) Production environment : sit with All the admins and tell them issue 

Thanks 

Suresh

On Friday, December 14, 2012 7:24:18 AM UTC+5:30, rahul sinha wrote:

Hello All,
  I had a question for all of you. I had to run a load test for a web based application and have set up a monitor using sitescope to monitor the CPU Utilization, memory and Unix Resources on the server. The issue i am facing is that the Websphere server  is shared by many apps and therefore even before beginning the load test the CPU utilization on the server would fluctuate between 50-80%. As expected, when i ran the load test i saw that the utilization was somewhere in the range of 86-92% throughout the test.

I advised the developer to increase the JVM memory size but this did not do the trick.
FYI: The counters for memory was 1% during the entire test and the Unix resources were also in the normal range.

Has anyone encountered a similar problem and can advise what might be the possible reason and solution to this problem.

Appreciate your help.

Thanks,
Rahul

--
 
 

[Vivek]

I Second that. Think twice before increasing the heap size to any value of your choice, as larger heap means more room for garbage collector to scan and mark dead objects to remove which results in increased time spent in GC and the effects could be detrimental.

Get yourself familiar with a little something called Ergonomics before you start to play around with the JVM startup parameters.

--