0
votes

When I run a JMeter Test Plan (jmx file) which comprises a single Thread Group that makes a SOAP request running 50 threads using a LoadRunner Controller the test runs perfectly, and I can achive  45 TPS consistently. This validates that the JMeter test plan and associated property files is uses are working perfectly under LoadRunner.

When I run the same test with JMeter the throughput drops over the course of the test run, reaching maybe 5 TPS after 15 minutes.

JMeter has maximum resources, and the developers have confirmend that the backend is processing as expected, the only difference seems to be how JMeter handles the requests, and how LoadRunner does as they use the same JMeter jmf file, data files, and property files for each test.

Can anybody suggest what the limiting factor can be for JMeter as LoadRunner runs the same test without suffering from decreasing throughut when everything else is identical test wise.

Thanks in advance,

NWC

1

1 Answers

0
votes

Well-behaved JMeter doesn't have any limitations, just make sure to:

  1. Follow JMeter Best Practices
  2. Make sure that JMeter has enough headroom to operate in terms of CPU, RAM, etc. as if there is a lack of hardware/software resources JMeter will not be able to send requests fast enough. If you don't have any better monitoring solution you can consider using JMeter PerfMon Plugin
  3. Pay attention to JVM metrics as well, ideally heap occupation should be between 40 and 70% otherwise you will suffer from too frequent/long GC events/chunking

Check out Concurrent, High Throughput Performance Testing with JMeter article for example JMeter-side performance bottlenecks analysis and JMeter tuning tips