[Grinder-use] Creating & Measuring Grinder Testcases
Distributed load testing framework - Java, Jython, or Clojure scripts.
Brought to you by:
philipa
|
From: <joh...@st...> - 2006-04-27 20:15:15
|
I would like to simulate an increasing number of (Virtual) User doing a special process (script) on the webapplication. In the peak the amount of virtual users should be 100.
The test case looks like:
Eg. I start with 1 single user and ramp up every 60 seconds a new user, so that after 100 minutes I have my 100 parallel users running their task.
After the test finished I would like to analyse how response time varies with the amount of concurrent virtual users. ( I know the rate of the increase of the users over the time, now I only need the response times).
My idea was to take a look at the concrete response times of every test performed only of the first worker and check if there are any decrease compared to the same tests done by that worker before.
(Do you think this is enough? Do you think the test like this will give helpfull data?)
Do you think it would be better, taking more machines. Some are running the load against webserver from 2-99 users. And one machine (the first user) is doing the request as a single user from another machine alone?
-------------------------------------------------------------------
Is the configuration of the property File correct like that?
grinder.processes = 100 (100 Workers at all)
grinder.threads = 1 (Only 1 Thread per Worker)
grinder.runs = 0 (Test should run vor ever, because the time of 100 minutes is important )
grinder.processIncrement = 1 (Ramp up one Worker each intervall = 60 seconds)
grinder.processIncrementInterval = 60000 (Every 60 seconds 1 worker is started)
grinder.initialProcesses = 1 (Start with one worker)
grinder.duration = 3600000 (60 min x 60 seconds x 1000, after 1 hour the test should stop.. than also all 100 worker have been started.)
-------------------------------------------------------------------
Best regards
jo
|