Load Runner 9.5 Flex script - needs realistic response time

I've successfully created a VU Gen script (Flex, Web http) that log in to a website. When I replay the script, even with Think time set to "as recorded", The Transaction duration (run time) is always much faster than if I manually perform the same click streams on the browser session, 17 sec as oppose to real wall clock time of 40, as an example. If I set to ignore Think time it would run even faster.

What Runtime settings can I set to get a more realistic user response time of performming the same functions from a browser? This problem compounds when I do a load test of multiple users for longer period. Worse of all, the Average Response time is completely misleading.
  • I had the same problem when load testing a flex application. The response time does not include the client side rendering time, which usually takes most of the time in case of applications using flex. The response time only includes the time till it recevies the last byte of data.

    There are two things you can do:

    1. Run the same scenario for a single user manually and capture the time for all screens to load completely. Do the same test using LR. Find out the difference between the two times. See if there is a pattern in the difference. If yes extrapolate your results based on the delta u observe. (This may not be the best/recommended way, so go for option 2)

    2. Use QTP scripts in LR (GUI Vuser). QTP also has local system montitoring capabilities which can give a breakdown of the time taken by the client.

  • Thanks for the suggestions. Since we don't have license to use QTP with LR, would have to take path 1 to get the differences.