Experiences with applying a genetic algorithm to determine an information systems architecture

While determining information systems architectures (ISA), business systems planning (BSP) is a well-known method to join processes and data classes to subsystems. BSP matrices have generally been rearranged without describing the underlying methods. Meanwhile, various techniques have been developed for solving the ISA problem. Since exact optimization methods often fail to provide results for large ISA problems, different heuristics have been applied. A new heuristic for solving the ISA problem is the application of genetic algorithms (GA). This paper examines the application of a simple GA to the ISA problem and compares the results of applying the GA with those obtained by exact methods.