Master-Slave Parallel Genetic Algorithm Framework on MPI

Genetic Algorithm (GA) is an effective tool to solve the problem of complex system optimization. Its implicit parallelism character gives itself a good prospect in parallel system frameworks. In this paper, we added the MPI into the “D.L. Carroll's FORTRAN Genetic Algorithm Driver”, and developed it into a reusable Master-Slave GA framework. To testify its validity, we used the standard GA test function running on a COW cluster consisting of two double Xeon workstations. This framework makes the use of Parallel GA more easily to users lack of the experience of parallel programming.