Model and Method for Optimizing Computational Processes in Parallel Computing Systems

详细

The paper proposes a model and a method for optimizing computational processes in parallel computing systems. The model is constructed by the decomposition of computer programs for solving problems into relatively independent elements (blocks) and presenting the information dependence of the blocks in the form of corresponding directed graphs. Here, the computational process organization consists in the dynamic allocation of resources of a computing system for the implementation of operations of each block. As an efficiency indicator of this process, we take the time of implementing operations of all blocks of considered programs (the solution time of related problems). The goal of the optimization is to reduce this time as much as possible. To form the optimal resource allocation, we propose the method based on the branch-and-bound procedure.