INCREASING THE EXECUTION SPEED OF DISCRETE EVENT SIMULATION LANGUAGES THROUGH THE USE OF ADVANCED EVENT FILE ALGORITHMS (COMPUTER).

摘要

Digital simulation is an important analytical tool for studying and optimizing the performance of complex systems. The major disadvantage associated with simulation is the expense incurred from lengthy computer runs. This research was directed at increasing the execution speed of discrete event simulation languages since they provide the implementation vehicle for many simulation models.;Much previous event file research had performed comparative tests of event file algorithms using artificially generated event streams. Another research method used "typical" simulation models as the source of test data. This research, however, employed a new approach which involved identifying experimental factors believed to influence execution speed. Simulation models were obtained to generate event streams exhibiting factor levels as needed for a scientifically designed, statistically valid and unbiased experiment.;Results of the experiment showed great potential for increasing execution speed in discrete event simulation languages by substituting one of the advanced algorithms tested for the current method. Blackstone's "two list" method and the fixed time increment indexed list method of Vaucher, Duvall, and Wyman were shown to be clearly superior to the alternative algorithms when compared on execution speed, robustness, simplicity, and other criteria.;In many simulation models, a large fraction (estimated to be as high as 80%) of total execution time is occupied by event file maintenance. Accordingly, this research has addressed increased efficiency in that aspect of simulation languages as the most promising avenue for increasing overall efficiency. The event file maintenance method most commonly used in current simulation languages was compared to nine alternative methods.