Throughput limitation HP LG versus JMeter

We have a very basic web script that has 2 requests.

When we run a test using JMeter, we're able to test to around 200TPS with 400 threads. However when we try to run the same test in HP (400 VU from the same host) the LG becomes a bottleneck at 80TPS due to a CPU constraint.

We've reduced also logging in HP and tried with different think times and increased iteration pacing, however this hasnt changed the result.

Does anyone have any ideas about this? Is it expected that HP VUs have a larger resource footprint than JMeter?

 

  • Hi Mason,

    JMeter users are very light weight and do the performance testing with some high amount of users for applications as user pass as thread and can accomidate like 400 users on with 200 TPS.

                But LoadRunner it is not possible because it executes the thread based upon the Load Genearator capacity which may be same (as JMeter test) ideally for LoadRunner it takes 2MB for each Vuser running for Web HTTP/HTML Protocol and consumes more for other protocols too like Flex, Truce Client..etc.

    It also depends upon the free space available for the Vusers on LG after the OS has been started. So it is not possible to get that as many users as JMeter with same TPS in a same machine for LoadRunner.

           It is not a drawback as LoadRunner supports different types of protocols but it is different for handling the users with Down , Pending ,Initaited and Ready and Running states to do more realistic end user behaviour on application.