Need help with StresStimulus? Start here.
The "request response time" measurement
|
Tony L. Posts: 8
6/1/2011
|
The measurement of “request response time” seems to be messed up by the client machine’s lagging time in handling many VUs and re-playing HTTP requests, due to its insufficient processing power.
My observation is: while VUs are increasing and since the point the local processor reaches 100%, the average response time increases drastically. However, the monitoring on the server side suggests the server is well. In such case, the measured response time does not reflect the real server performance. It is merely the congestion in the client side.
|
|
|
0
link
|
|
Vadim @StresStimulus Administrator Posts: 583
6/1/2011
|
Tony,
This is a correct observation, as client approaches 100% CPU utilization, slowness will affect accuracy of response time measurement. According to this source, when CPU utilization is reaching 80-85% it is recommended to use more client agents. In the current version of StresStimulus you can try using more clients to support desired number of VUs.
Since this is issue are not specific to StresStimulus I will move this post to Support Questions forum
Cheers,
-Vadim
|
|
|
0
link
|
|
Tony L. Posts: 8
6/1/2011
|
thanks for the confirmation.
i am thinking it might be helpful if the program can be smart enough to "tell", for a defined set of test cases, the number of VUs it's capable to handle, by analysing a test run, which sounds like to 'stress run' the client.
|
|
|
0
link
|
|
Vadim @StresStimulus Administrator Posts: 583
6/1/2011
|
I like the name - 'stress run' the client. I will check if other tools can determine the max number of VUs to see what we can learn.
|
|
|
0
link
|
Copyright © 2024 Stimulus Technology