A typical model for distributed computing is to have a main program thread that runs on one processor. This thread spawns a number of tasks from time-to-time. When tasks are spawned, they are sent to other processors for completion and the main thread waits till the results of all task are received from the remote processors. This is the typical fork join paradigm. This paradigm results in an interesting scheduling problem that is studied in the paper. Several heuristics are proposed for various variants of this scheduling problem.