Abstract

Some discrete simulation models are too large to be executed on a single processor; in other cases, results might be required faster than a sequential execution can provide them. Such models are candidates for parallelization. Here, models are distributed among several processors, and are then executed with careful synchronization.
This paper provides an introduction to the fundamentals and methods of the parallel execution of simulation models, with a focus on model-based parallelization. The paper describes the two main classes of parallel simulation methods, conservative and optimistic simulation, their respective advantages and shortcomings. A second focus is put on static and dynamic load balancing, with a dynamic load balancing method first developed to accelerate the simulation of transportation systems being introduced in some detail. In addition, the paper describes some typical applications of model-based parallelization.