Think you found a bug? Report it here.
Mix Weight does not act as expected given the description
Brianna Blanchard Posts: 76
2/19/2014

The description is Mix Weight is "The relative frequency of the Test Case replays in the mix. Every VU is assigned to a specific Test Case selected in a roundrobin order, while skipping some of them to achieve the VU distribution corresponding to the mix weights."
However, if you have a number of users that is a smaller number than the total mix weights, say, 50 users and mix weights of 75 and 25, the test cases will get an even split of the 50 users. Given the words "relative frequency", I expected entering 75 and 25 to give a 75%, 25% split, even with only 50 users. If I use 100 or more users, the mix weights acts as I expected.
I was using 75 and 25 rather than 3 and 1 to make my automated data processing easier.


0
link

Max @StresStimulus Administrator Posts: 101
2/19/2014

Hi, Thanks for your comment. We will adjust the documentation to reflect that "relative frequency" means units, not percentage. As you noted, your case, mix weights 75 and 25 will not work, but 3 and 1 will work. This is because StresStimulus uses a roundrobin algorithm to assign test cases to VUs given the test case mix weight. Therefore the sum of the mix weights should be less or equal to the number of VUs. Also, we will also make an improvement in this area: that will support 75 / 25 ration. StresStimulus will try to normalize mix weights internally by dividing them by the greatest common divisor, so 75/25 will work just good as 3/1. We will release a new build this week.


0
link

Max @StresStimulus Administrator Posts: 101
2/19/2014

The improvement to support percents (when possible) in the test case mix weights and the documentation update is included into v3.5 update 4 released today.
edited by admin on 12/29/2016


0
link

Brianna Blanchard Posts: 76
2/19/2014

Great, thanks for the prompt turnaround.


0
link

Copyright © 2018 Stimulus Technology