PI ACE 2011 perform really better than previous version : it start faster, schedule better and then report very few skipped calculations compared to ACE 2.1.32 ! That's great but... in case of burst of input data and natural scheduling it reachs much faster the 2 GBytes limit of working set of windows process ! And then the corresponding module is stale.
I already reduced the number of contexts of such module (from 3000 to 1000 in version 2.1.32 and recently from 1000 to 500 in version 2.1.50) to avoid such problem. But it still occurs from time to time. What could be the good management of queuing ? is there a way to extent those queues to files (as it is done for event queue on Pi server side) ? Can i enforce using ACE code the skipping of calculation (better than the killing of a module) ?