I have to modify and use parallel processing for a large and complex system which is implemented in Perl. Each child process or threads are supposed to use memory around 1 GB and will run for around 1-3 hours on Solaris Server. Total parent's execution time would be around 12-14 hours.
My question is - what is your recommendation - use of ithreads or multi processing using fork(). 100s of processes or threads need to be created with a concurrency limit of around 15 at a time to loop over list of 130-150 jobs. Considering amount of memory it is going to use and execution duration, which approach you would suggest? Relibility of the overall processing is important since it will run in production environment.