5.1 Introduction

A distributed simulation of a dynamic system consists of a group of component or subsystem simulations that communicate across a network during execution. The individual simulations may run on a single computer system or they may run on computers that are separated by large distances. All of the simulations run in parallel and pass data among themselves as needed over the network. The techniques of distributed simulation are applicable to both non-real-time simulations and to real-time simulation applications.

For non-real-time simulation applications, a primary benefit of distributed simulation is the execution speedup that results from running on multiple processors in parallel. This assumes that the additional overhead required by the network communication is small in comparison to the execution time saved by running the component simulations in parallel. In a non-real-time application, there should be no difference in the results of the distributed simulation when compared to the results of an equivalent system simulation that runs as a single process on one computer. To achieve this result, there can be no errors or lost data packets during network communication and the simulations must always maintain proper time synchronization among themselves. Although the issues involved are nontrivial, it is possible to satisfy these requirements with the use of appropriate communication methods. These techniques will be discussed later in this chapter.

In real-time simulation applications, we can also achieve the execution speedup benefits of running the subsystem simulations in parallel. Models can be implemented that are more complex...