We have implemented a system enabling to approximate a solution of a given problem with a method of parallel genetic algorithm. Our system is prepared to solve a generic problem defined by the user. We have performed the performance test of the system (see section 4.1). It seems that the parallel computation of genetic algorithm performed by our system kernel achieves better results with increasing number of processors. Of course the system is handicapped by its versatility. It is obvious that a genetic algorithm system implemented only for optimization of a specific problem can achieve better results than our system.
Our objective was to design an open system. Hence we have created the system structure modular and we have prepared it for the further development (see section 4.2 for details). Mainly is the system prepared for incorporating of new parallelization methods. We have defined all kernel interfaces, so any system component can be extended or a new one can be added to the system package.