Increasingly, computer simulation is directed at predicting the behavior and performance of large manmade systems that themselves include multiple instances of imbedded digital computers. Often the computers are distributed, sometimes over wide geographic distances, and the system modelling becomes largely a combination computer/communication simulation. The type of simulation needed here can be characterized as having some elements that are simulated in a conventional sense where a statistical or descriptive model of the element is used. But other elements, particularly the imbedded computers, are emulated, which is to say that the computations performed nearly duplicate the functionality of that real-world element. For example, a ``tracker'' really tracks. It does not simply provide results that are in conformance with how a tracker should track.
In this manner, the simulator becomes both a predictor of system performance and an active participant in the system development as the behavior of the emulated elements is refined and evolved.
In 1987, the Mark III Hypercube Applications group at JPL undertook the most computationally demanding simulation of this type yet proposed: the detailed simulation of the global Strategic Defense Initiative System-sometimes known as Star Wars [Meier:89a;90a], [Warren:88a], [Zimmerman:89a].
A parallel processor was chosen to perform the simulation both because of its ability to deliver the computational power required and because it was closely reflective of the class of machines that might be used for the imbedded computers in the SDI System-that is, the simulation was helping to prove the applicability of parallel processing for complex real-time system applications. The Mark IIIfp Hypercube was the host machine of choice at the time (1987-1989).