For the past few months I have been struggling with the mysteriously poor performing QuickTest Professional (QTP) on Windows. When tested similar transactions using the LoadRunner (LR) tool, the average response times were satisfactory -- mostly under 2 seconds. However when tested with QTP, the averages were unusually high - they are mostly between 2 and 4 seconds. LoadRunner data eliminated the web and application servers from the picture. Because the LR reported averages are in the expected range, the web and app servers might be performing well. Hence there is no apparent reason to suspect web/app servers for the slowness observed through the QTP tool. That leaves the QTP client as the primary suspect.
There is no improvement in the average transaction response times even after downgrading the version of QTP tool and the Internet Explorer web browser. Luckily one of my colleagues at work noticed slight improvement to the averages when we are not monitoring the QTP test running through our VNC viewers. The response times were better when we either close the VNC viewer window completely or minimize it. That was an interesting observation. VNC service appears to be interfering with the web browser's rendering performance; and due to that the recorded response times were poor. So I stopped the VNC server/service on the QTP client machine; and connected to the QTP client through the "Remote Desktop connection" from another PC. Finally I re-ran the same QTP test that I ran earlier when I was connecting through the VNC. Voila! the QTP response times are almost coinciding with the LoadRunner reported response times.
To make it more attractive, here is some sample data that shows the difference between the response times when connected through the VNC and the Remote Desktop connections:
----------------------------------------------------
VNC Remote
Viewer Desktop
----------------------------------------------------
Average Logon 2.59 sec 0.76 sec
Avg. Page Load 2.91 sec 0.77 sec
Avg. Page Save 3.82 sec 2.05 sec
----------------------------------------------------
Moral of the story:
When running performance tests with QuckTest Pro (QTP) tool on remote clients, consider connecting to the remote client using "Remote Desktop" connection. Web browser might perform suboptimally under the VNC, which might skew the overall performance data.
And of course, I do not have any evidence other than the above data that proves VNC guilty. It was just my observation.
________________
Technorati Tags:
QTP |
QuickTest Pro |
LoadRunner |
VNC |
Remote Desktop |
RDP |
Windows