Extracting a memory snapshot is pretty fast (probably 5 seconds) even to a large memory.
Probably analyzing the files will take some time at the first time, because I need to sort the files using plain old quick sort, what will take some time. I have tried using a database and we got even slower responses I perform a single query for each object selected. The best algorithm I found was using a sorted file.
So, the only thing is, you will have to select some spots to extract the memory snapshots.
The analysis takes place on in webapp, no? So this won't impact the testsuite.
I think by default we will just extract the memory snapshots at the end of each server run, as a part of the server:stop functionality.
Yes.. the analysis will take place in the webapp.
Anyway, as the profiler is dependent on a native DLL, we will have to make this an optional feature on the testsuite.
And also, remember that JVMTI (the library I used to capture the memory snapshots) is only available at Java5.