Parallel computing in application to global optimization problem solving

This paper presents the results of a research of parallelization possibility of different optimization methods used in finding the global extremum. A parallel algorithm for finding the global extremum based on modified Box's complex method with explicit and implicit constraints is proposed. We also determined the optimal number of computing nodes needed to provide predefined calculation accuracy and computational stability. The results of the parallel adaptation of the multi-extremal objective function global minimum finding algorithm based on simulated annealing method are presented. The reliability of finding the global minimum, depending on the number of nodes used in parallel computing system is investigated. We showed that parallel version of the simulated annealing method makes it possible to reliably find the area of the global minimum in a small amount of time. Also the option of using genetic algorithms with parallel computing systems for global optimization problem solving is de scribed. A comparative assessment of effectiveness of these methods is made and the recommendations for their use in different tasks are given.