Efficient implementation of multiprocessor scheduling algorithms on a simulation testbed

A layered and modular approach to implementing a process-oriented simulator testbed is described. The simulation kernel is supported by a threads library and is, in turn, capable of supporting distinct domains or application areas for various applications. The testbed offers an implementation methodology for testing novel simulation algorithms at the domain level, without interfacing with the kernel. To demonstrate its utility, a novel algorithm for simulating multiprocessing with round-robin scheduling is presented. The algorithm is more complex than the naive round-robin implementation in use, but offers significant performance improvement. Copyright (C) 2004 John Wiley Sons, Ltd.