AN ASYNCHRONOUS MESSAGE BASED APPROACH FOR DISTRIBUTED DISCRETE EVENT SIMULATION

Abstract

Traditional discrete event simulations employ an inherently sequential simulation algorithm. In practice, simulation of large and complex systems may take prohibitively large time to execute on a uniprocessor system, If simulated using this approach. This is because of the sequentiality of the approach wherein only one event is simulated at a time. Moreover only a modest number of events can be simulated. This severely limits the use of this approach for large systems. To speed up the simulation, the sequential approach has to be discarded. This dissertation aims at demonstrating such an approach, called, "the asynchronous message based approach", that performs better in time. Basically, this is a distributed discrete event simulation program that executes on a set of processors communicating asynchronously through message passing. The basic distributed simulation scheme, which uses time encoding, is described. Its major shortcoming is the possibility of deadlock. Its avoidance through null messages is shown. The distributed simulation scheme is applied to a real system (CSMA/CD) and its results are discussed. An environment to write distributed simulation programs has also been developed in the work.