Research On Mechanisms For Resource Management In Peer-to-Peer Networks

Peer-to-Peer (P2P) is a most important inherent characteristic of the Internet. Today, P2P applications have become the overwhelming network services attracting a huge number of users. However, some serious problems on resource management of P2P networks have become the obstacle in the development of the P2P networks.In this dissertation, we focus on the problem of resource management of P2P networks, based on Incomplete Information Theory and Probability Theory. The main fruits are described as follow:(1) Based on the concept of decentralized grouping, a new architecture of P2P network resource discovery is proposed, which incorporates the underlying locality into the semantic overlays. The query and the major management operations are processed in the group one by one. In this way, the globe status maintenance can be avoided, and the performance of query procedure has been enhanced largely. The simulation results have validated the effectiveness of this mechanism, including average diameter, average management overhead, and average searching overhead.(2) Based on the analysis of the "free-rider" problem of file sharing in P2P networks, an incentive mechanism is designed to induce all the users to share the resource as much as possible according to his true capability. Using this mechanism, the users can be inspired to provide commons by taxing them. This mechanism can be also used to manage the P2P networks. Finally, the simulation results have validated the effectiveness of this mechanism.(3) The user's bandwidth can be regarded as private goods. Based on this idea and the theory of the Bayesian game of incomplete information, a new incentive mechanism is proposed to encourage all the users to report the true type of access-link that they really use. This new mechanism also provides the different types of users with corresponding services which can achieve the rational fairness among the users. The effectiveness of this mechanism has been validated by the simulation results. (4) A new probability-based adaptive initial reputation mechanism for the P2P networks is proposed to exclude the whitewasher, which reenters the system by changing a new ID after bad behaviors. To avoid the system fluctuating for actions of few whitewashers, a new reputation mechanism is incorporated with group-based architecture, which can localize the impact of whitewashers in their own groups. The simulation results have validated the effectiveness of this reputation mechanism.