Abstract
The parallel evolutionary P2P networking technique, which we previously proposed, divides all nodes composing a large-scale P2P network into multiple node groups, and then reconstructs the network topologies by an evolutionary algorithm in each node group. In this technique, a timing at which a set of network topologies (a population of the evolutionary algorithm) is evaluated and then reconstructed by evolutionary operators is the same for all of the node groups. However, this simultaneous evaluation and topology reconstruction has been shown to be not effective in terms of adaptation of the network topologies to users' (nodes') demands in our previous study. Therefore, in the present paper we propose a method that makes all the node groups evaluate and reconstruct a set of network topologies serially. Simulation results show that the proposed method indeed improves the performance of the parallel evolutionary P2P networking technique.
