Automatic Generation of Real-Time Aircraft Simulation System Configurations

Building configurations for real-time aircraft simulation systems is a challenging task. It involves the distribution of the applications among different scheduling processes, bound to different CPU's, in such a way that the applications' priority and expected execution order are taken into account. In this thesis, we report on a study conducted at CAE Inc., a world leading manufacturer of flight simulation products, in which we have developed an approach to automatically build configurations. Our approach is based on a greedy algorithm that uses heuristics to distribute many applications into different partitions in such a way that inter-partition communication is minimized, the load across partitions is balanced, and each partition is denoted as a binary tree (the data structure used by the scheduler to run the applications). The configuration is also constrained by the priority and execution time of the applications. When applied to CAE, our approach produces configurations that in most cases outperform or are similar to those generated by a domain expert