Continuous rampUsers until the hit rate is capted or while a chosen duration average is below a given limit ?


I wonder how to design a test scenario that will run on various kind of servers without any changes and will give us what are the chosen platform limits. In fact I don’t want to give any value to the ramp users feature instead I wonder how to write a generic custom smart ramp users feature that will be able to :

  • add quickly new users if the hit rate grows quickly
  • slow down new users injection if the hit rate is slowing down
  • stabilize users count if the hit rate couldn’t increase more and/or a chosen period moving average is increasing “too” fast.
  • may be also remove some users to return back to previously seen better performances
  • keep the selected number of users for the specified amount of time

setUp(scn.inject(rampUsersSmartFindLimit over(10 minutes))).protocols(httpConf)

=> stabilized after 7 minutes with 3400 users, reached 4200 hit/s

I guess that this mechanism will have to be “smart” enough in order to take into account that the test platform may require an unknown warmup period in order to achieve its best performances.

Has anyone work on such feature ? any hints on how to code this ?

As is, the inject API doesn’t have access to heuristics.

Then, I’m not sure I see your point: you’re making the load compensate the system under test performance.
Looks like intended coordinated omission.