A novel selection approach for replicated multicast servers using genetic algorithm

Multicast server replication is an effective technique for multicast to enhance the performance and scalability. It employs multiple replicated servers located in different places to deliver information in one multicast service. Server selection is the primary issue of multicast server replication; however, to achieve better throughput and fairness among clients, it is not a trivial problem to assign servers to large amount of clients. In this paper, we propose a novel approach using genetic algorithm to handle the NP-hard replicated multicast servers selection (RMSS). A two-level coding scheme is developed to map RMSS solution space into the chromosome space, which is believed of great importance to the whole approach. In this step, a method utilizing Dijkstra algorithm and random disturbance is designed to create a random multicast tree guaranteeing loop freedom and connectivity. Fitness function and genetic operators are also discussed. The simulations clearly demonstrate that our proposed genetic algorithm outperforms the heuristics in the sense of yielding high-quality solutions. In addition, it also shows better performance in load balancing.