Traffic prediction-based routing algorithm over structured P2P networks

Abstract

Routing has always been a great challenge for structured Peer-to-Peer (P2P) networks. There are a lot of representative structured routing algorithms for P2P networks, but these algorithms do not guarantee the quality of service (QoS) for real-time P2P applications. Addressing this challenge, a traffic prediction-based structured routing algorithm over P2P networks (TPSR) is proposed. Our contributions are described as below. We firstly analyze P2P traffic features and then build a wavelet neural-network predicting model. Secondly, we employ the traffic prediction model to predict the future state of each peer, such as normal or congestion, and let each peer update its routing table. In this way the requesting peers always get a resource list which contains the best resource peers. Simulation results demonstrate that TPSR has higher transmission success rate and lower end to end delay than other structured routing algorithms. Thus, TPSR can guarantee the QoS for real-time P2P applications.