Browse Reports

Simulations of computer systems have traditionally been performed on a single sequential computer, even if the system to be simulated contains a number of components which operate concurrently. An alternative would be to simulate these systems on a network of processors. With this approach, each processor would simulate one component of the system, hence the component simulations would proceed concurrently. By exploiting the modularity and concurrency in the system to be simulated, the simulation would itself be modular and concurrent. An accurate simulation must model the time behavior of the system as well as its input-output behavior. In order to avoid real-time constraints on the processors and communication network in the simulation facility, the simulation timing must use a time-independent algorithm. That is, the simulated behavior of each component should not depend on the speed at which the simulation is performed. With this time-independent approach, additional coordination operations are required to prevent deadlock of the simulation. This coordination can be provided without any centralized control. Instead, the program for the simulation of each component is modified t simulations. Additional termination operations are also required to assure that the simulation will terminate under the exact same conditions that the system being simulated would terminate. These operations can also be provided without any centralization of control or real-time constraints. Furthermore a simulation which uses these coordination and termination operations is probably correct. That is, the simulation will accurately model both the time behavior and the input-output behavior of the system.