We are running a load test with one Web/HTTP script and 25 user load. The response times for all the transactions gradually increases from 5 seconds during ramp up to 90+ seconds during steady state. But the resource utilization at Application layer is below 10% and Load generator is below 20%. There is no saturation observed in CPU, Memory, Disk and Threads utilization.
During steady state when we run the Vugen script in Load generator machine, the response time is below 5 seconds. Also, When we manually run the workflow in browser, the response time is below 5 seconds. logging is disabled in script. Think times are initialized outside transactions. Pacing is set to 30 seconds.
Based on above observations we think the issue with the Load Generator.
We found 2 solutions to resolve the issue:
1) Changing the multithreading option from Thread to Process in script run time settings. The issue still occurs when we set it to Process but the probability is 20%. When option is set to Thread, the issue occurs all the time.
2) Use of two load generators instead of one. But the issue still occurs with a 10%-20% probability when we use this option.
We are still not able to figure out the exact root cause of the issue. Please let me know for more information.
Vugen version: 12.55 Build 0911
HPE PC: 12.55 Build 12.55.3489.0
Load Generator config: VM based. Windows Server 2012 R2 64 bit, Xeon V2 4 CPU at 2.8 GHZ, 16 GB Memory.
Thanks in advance.