2

I was doing performance testing on an application in my organization with load of 250 concurrent users with ramp up of 50,100 then 250. I am using IBM Rational performance tester(RPT). I have used IE browser for recording the scripts.

The pattern which I have observed is that when 250 users were running for some time(nearly 14 minutes) after ramp up stage, the response time graph decreased sharply. When I check the application under test on browser it showed "Page cannot be displayed". After some time the response graph again began to rise and when I check the application under test on browser it became up again. This behavior was occurring in repeated manner. I asked application server handling team whether they have restarted the application but they told me that they have not done so.

The application under test was becoming up on its own.

The application server handling team hinted that this might be happening due to the maximum limit which server can handle for concurrent requests. But they also told me that they have not set any limit. They are using Oracle BEA WEBLOGIC SERVER.

Can anyone here provide me any guidance about exact reason of this behavior. Specially I am concerned with that how server is becoming up again and again without anybody restarting it?

3

Load-testing without monitoring is worthless. Load-testing is about finding bottlenecks in your infrastructure. So you NEED to monitor your infrastructure closely while doing a load-test.

There could be a lot of reasons why the server stopped responding and started responding again. Some examples:

  • Server out of memory
  • Server or Client default limits, also operating systems have limits of connections
  • Denial of service protection in the software (seen this one a couple of times)
  • Network firewall/router limiting traffic
  • Application crashed and has an auto restart
  • CPU became so hot the hardware rebooted, heat protection

I would advice to execute the load-test again, but now combined with someone from the application handling team and some good monitoring and logging tools. Then analyze not only the results from the response times, but also how it affects your infrastructure.

Other reads:

2

Looks like a "classic" memory leak or excessive work of the Garbage Collector

Try monitoring JVM heap via JMX Console during the test, I believe you will see memory growing up till the point when application stops responding and intensive garbage collection. If it is the case - some JVM tuning will be required.

Also pay attention to overall server health metrics, i.e. RAM usage, CPU usage, Disk and Network IO as it may occur due to lack of free RAM. You can do it using i.e. JMeter PerfMon Plugin

1

It looks like a memory leak issue to me. Like the guys above had mentioned already, monitor the server level activities and tune the application accordingly. Also, I doubt the application's capability of handling 250 users. I would advice to have a check on the architecture and the increase the server capacity.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.