Which implementation will have lesser impact on performance......
Discussion posts and replies are publicly visible
Generally I'd rather implement something closer to your first approach (split process flow with AND node, call subprocess synchronously, get data from it when done), because while the Send Message approach can work, it can be a nightmare in terms of management, debugging and upkeep. I don't expect either approach would be all that different in terms of performance unless the subprocess is doing something unusually process-intensive.
thanks for your response Mike....
The sub-process which we are referring can last upto 100hours(worst case ),which will do polling to external system(integration) for every 45 mins.
Would you think this process-intensive?