9523111 Andradottir The research objective is to investigate how the efficiency of simulation can be improved by conducting the simulation experiment on parallel processors. The parallel simulation approach envisioned in the research involves segmenting the time horizon allocating processor to simulate the entire system over a fraction of the total time horizon. This is equivalent to allocating only a portion of the total time segments to a processor. A major difficulty in applying time segmentation to simulate stochastic systems on parallel a processor is finding efficient ways of generating a valid sample path of the underlying system. This is complicated by the fact that the initial conditions for the simulation on all but the first processor or time segment are unknown when the simulation is started. The new approach in this research is concerned with determining both the scope of applicability and the efficiency of a parallel simulation approach. The investigation will determine when parallel simulation approachs can be used to efficiently analyze complex systems. Simulation is a key tool for studying and analyzing complex and dynamic systems. Its application domain is large and includes engineering, economic, and social systems. One common criticism in the use of simulation is the amount of computational effort required before meaningful data can be generated. The parallel simulation approach, investigated in this research, is a move towards speeding up the simulation process on computers. If successful, the technique will see immediate application and incorporation into commercial simulation software. Large scale application will produce significant savings in time and money and shorten the length of time required to perform a simulation study.
