Simulation study on the architecture optimization and task scheduling of a parallel flight training simulator

A parallel flight training simulator technique is being developed at Beijing University of Aeronautics and Astronautics. At the first stage, a simulation environment is set up to optimize the task scheduling and evaluate the multiprocessor architectures. One main purpose of this paper is the choice of a proper multiprocessor architecture for the simulator systems by simulation. Processors are assumed to cooperate in a message passing fashion, and the messages are exchanged through common memory areas. Four multiple bus multiprocessor architectures are considered, which differ in the location of the common memory modules. Another main purpose of this paper is to explore an effective task scheduling. Using the simulation environment of the parallel computer system used in the simulator, the quasi-dynamic task schedule method is presented. This task scheduling is dynamically performed in the simulation environment so as to obtain a static task scheduling in which the dynamic execution of task is considered. Just like dynamic task scheduling, our task schedule strategy can obtain very effective schedulings because the effect of dynamic conditions, such as the system synchronization, the contention for shared resources, the communication between processors and so on, can be included in the scheduling. On the other hand, the scheduling is obtained by off-line, which has the benefits of static task scheduling.