clicker wrote:What's the criteria for a test being "loadtest safe"?
As you know, the eValid approach for server loading is to run multiple eValid browsers all in parallel on one or more machines.
The only limitations on this idea -- other than the capacity of the machine itself -- is that there are two unitary resources involved: the browser cache and the desktop.
The browser cache is handled by eValid by having simply disallowing any cache operations. This means that individual tests can't take advantage of reloading pages from the cache -- they have to be re-downloaded from the server. But this is a conservative assumption anyway, and only serves to make the load test itself slightly heavier than it would be otherwise.
The desktop resource is more subtle: if two [or more] eValid instances try to ready something from the desktop there is a likelihood (some in testing will say this is a near certainty!) that the two instances will interfere with each other.
The way to provent this is to make sure that each testscript in the load scenario is "desktop safe", meaning that it does NOT require ANY actions on the desktop. A way to confirm "desktop safe" is to run the test with the eValid browser minimzed: if the test runs successfully that it is desktop safe!
So, really, "loadtest safe" is really another way of saying "desktop safe".
-- The eValid Support Team
by clicker » Thu Jan 21, 2010 3:23 pm