Implementation of a parallel Genetic Algorithm on a cluster of workstations: Traveling Salesman Problem, a case study

A parallel version of a Genetic Algorithm (GA) is presented and implemented on a cluster of workstations. Even though our algorithm is general enough to be applied to a wide variety of problems, we used it to obtain optimal and/or suboptimal solutions to the well-known Traveling Salesman Problem. The proposed algorithm is implemented using the Parallel Virtual Machine (PVM) library over a network of workstations. A master–slave paradigm is used to implement the proposed parallel/distributed Genetic Algorithm (PDGA), which is based on a distributed-memory approach. Tests were performed with clusters of 1, 2, 4, 8, and 16 workstations, using several real problems and population sizes. Results are presented to show how the performance of the algorithm is affected by variations on the number of slaves, population size, mutation rate, and mutation interval. The results presented show the utility, versatility, efficiency and potential value of the proposed parallel and distributed Genetic Algorithm to tackle NP-complete problems of the same nature.