Difference in Reponse time in TruClient & HTTP/HTML protocol
On execution of LoadRunner script I am facing the following issues:
I have created a script using TruClient-IE protocol for one of our release which is on ext-js. After executing the same on controller the response time is drastically high as compared to what I am getting using HTTP/HTML protocol on another release which is on XML on the same transactions or even when I am executing the same transactions manually.
With TruClient protocol it is as high as 16.99 seconds with 1 Vuser whereas when the same is done with HTTP/HTML protocol it is 3.183 seconds with 100 Vusers. Even when I am verifying the same manually it is not more than 3 seconds for the same test case. In both the cases think time/ wait time is same.
Also while executing, my Controller & Load Generator are two different machines.
Also when I scripted gmail login the response time with TruClient is 9 whereas with HTTP/HTML protocol the response time is 0.4 for the same transactions.
Please help me to know if I am missing something or if its an issue with Protocol.
1. Don't use automated tools like LoadRunner against third party services such as GMAIL. You are violating both Google's terms of service and the license of LoadRunner when you do this. It is akin to driving around a neighborhood and shooting at cars and homes you don't own.
3. Monitor your load generator to see what is impinged from a resource perspective. Use the same skills you use to find and isolate resource bottlenecks in your application to do the same for your load generators. Once you understand what resource is under stress when using your particular scripts then you will also understand what needs to happen to address this issue. For instance, if you have exceptional logging enabled in your TruClient but not in your HTTP virtual user, then consider disabling the additional logging which showed up as a high amount of disk activity in your monitoring.
4. Does it work for one user running? If not, then begin your troubleshooting here and go no further for integration into a performance test. For if it does not work for one, it will never scale for many.