ratio internals
Rick Lanham
Posts: 2
I have a simple web-site - 3 independent pages - page1, page2 & page3
I wrote a load with 3 simple actors (get1, get2, get3), each one waiting 400ms and then getting one copy of the corresponding page.
When I run this load for 1 minute with ratios 1:1:1 and 3 virtual clients the pages are retrieved in approximately equal amounts 151:142:142
When I run this load for 1 minute with ratios 1:1:1 and 5 virtual clients the pages are retrieved disproportionately 289:273:147
This (and other postings in your forum) suggest that each virtual client is allocated one actor (and only runs that actor) and that the allocation is based on the ratios. Is this assumption correct?
I wrote a load with 3 simple actors (get1, get2, get3), each one waiting 400ms and then getting one copy of the corresponding page.
When I run this load for 1 minute with ratios 1:1:1 and 3 virtual clients the pages are retrieved in approximately equal amounts 151:142:142
When I run this load for 1 minute with ratios 1:1:1 and 5 virtual clients the pages are retrieved disproportionately 289:273:147
This (and other postings in your forum) suggest that each virtual client is allocated one actor (and only runs that actor) and that the allocation is based on the ratios. Is this assumption correct?
This discussion has been closed.
Comments
That's the way it should work -- ANTS Load starts a thread with a VirtualClient class and populates its' Run() method with a script that it will loop through over and over again. It also assigns a cookie store and a line of state information to the thread so that each virtual client will work independently. If you have more users in the test than there are scripts, this is when the ratio becomes important, I think.
You'd really need a whole lot more virtual users to see the ratio being implemented properly.