Parallel Genetic Algorithm for Shortest Path Routing Problem with Collaborative Neighbors

Abstract

Shortest path routing is generally known as a kind of routing widely availed in computer networks nowadays. Although advantageous algorithms exist for finding the shortest path, however alternative methods may have their own supremacy. In this paper, parallel genetic algorithm for finding the shortest path routing is resorted to. In order to improve the computation time in this routing algorithm and to distribute the load balance between the processors as well, Fine-Grained parallel GA model is opted for. The proposed algorithm was simulated on Wraparound Mesh network topologies in different sizes. To this end, several experiments were anchored to identify the most influential parameters such as Migration rate, Mutation rate, and Crossover rate. The simulation result shows that best result of mutation rate is: about 0.02 and 0.03, and migration rate for transmission to the neighbor’s node is 3 of the best chromosomes. This study has already shown that through using performance-based GA which uses fine-grained parallel algorithms, timing germane shortest path routing can be improved.