| || |
Arrival rate & Markov question
What does load testing by "arrival rate" of the virtual users really mean in practice?
How can Markov chains be used in designing the load for performance testing(for example web site load testing)?
Re: Arrival rate & Markov question
Testing by "Arrival rate" in your situation sound the same as throughput. It is the rate at which requests for work arrive. Testing with this parameter would mean increasing your throughput by adding more virtual users. You can also increase throughput by having the virtual users work faster (i.e. decrease think and delay times).
I have also thought about using a Markov chain for test case design. The problem with doing that is, by definition a MC state transition from page to page must be independant of previous transitions, if you can estimate the chance of transiitioning from one page to the next, for all pages, you could use a Markov chain to get the probability of users being in a certain state. Which could then be used to weight your virtual users.