We use WinRunner as a load testing tool operating at the GUI layer as it allows us to record and playback user actions from a vast variety of user applications as if a real user had manually executed those actions.  We use WinRunner in addition to LoadRunner when we want to record user experience response time.  Visit mercuryinteractive.com/products/WinRunner/ for detailed information on WinRunner.

The following diagram shows how a GUI layer testing tool, such as WinRunner, operates. 

Diagram of how a GUI Level testing tool - such as WinRunner operates.

This is a diagram of a Protocol Level tool. NOTE the protocol tool does not see the events entered by mouse / KB etc. Click here to see more details on a protocol level tool.The WinRunner screen print shown below simulates a user starting up an Internet Explorer session and connecting to www.google.com.au before performing a Google search on the text: "Mercury Interactive".  As can be seen, WinRunner records each of the actions that the user performed on the desktop to get to and search the Google web site.  This is in contrast to the way that VUGen records the protocol that the client application generates.  However, both tools have their part to play in a load test.

The screen image below is a script example of how WinRunner recorded the events on the windows desktop to "Press Start" and then Invoke Internet Explorer by selecting the option marked "Internet".   The text "google.com" was recorded as being entered as a URL and the "return" key (<kReturn>) was then recorded so the IE loaded the Google site into the browser window.  The characters "Mercury Interactive" were then recorded as they were typed into the Google search field, followed by another <kReturn> to initiate the search.

Script Example of WinRunner recording a user searching the google web site.

As can be seen from this script example, WinRunner does nothing at the protocol layer (like VUGen would) but records and plays back user events, so that the underlying application operates as if a person was sitting at the desktop.

For WinRunner to operate, it needs to be in control of the PC, so that it can execute the user actions that had been previously recorded.  This is why one can not execute a load test with WinRunner as the means of load generation.  In order to simulate 100 users, one would need 100 PCs with WinRunner on each PC.

However, WinRunner is a valuable piece of load testing technology when used properly in a load test as it is the only means of determining the actual user response time, taking into account the processing that is executed on the clients hardware.  (As VUGen operates at a protocol level it is only able to measure at a protocol level.)

Please visit performance tests and network sensitivity tests for other testing situations where it is very appropriate to use WinRunner.  By using WinRunner in these situations, WinRunner usage will be extended beyond automated functional testing, increasing it's value to your testing team and organization.