Modeling process scheduling and system software in multiprocessors

This thesis presents new techniques for simulating multiprogramming and interrupt servicing in an execution driven simulation environment. We have incorporated these techniques into the Rice Parallel Processing Testbed (RPPT), a powerful tool for simulating and analyzing parallel computer systems and algorithms. A new approach for modeling system software overheads on the RPPT is also described. As part of this study, we validate the RPPT against a distributed system running the V operating system and outline possible applications of the system. The model of the distributed system makes use of the new interrupt and timeslicing capabilities of RPPT, and the capability to assign priorities to processes. The implementation of a separate operating system module for the architecture is shown to be effective in the correct simulation of the software overheads and in the logical division of the simulation environment