A cluster based incentive mechanism for P2P systems

Peer-to-Peer (P2P) networking is distributed computing paradigm, in which the nodes are self-organized and can directly exploit resources from each other without dedicated servers. Free riders in. the Peer-to-Peer systems are the nodes that only consume services but provide little or nothing in return. They seriously degrade the fault-tolerance, scalability and content availability of the Peer-to-Peer systems. The solution to this problem in Peer-to-Peer systems is to have incentive mechanisms that aim to improve the network utility by influencing the nodes to become more cooperative. This thesis proposes seven design requirements for an incentive mechanism according to the characteristics of Peer-to-Peer systems, latest distributed computing development trends and the related implementation techniques. This thesis also provides a classification of the existing incentive mechanisms for Peer-to-Peer systems. For each category, the thesis, outlines their principle, provides typical examples, applications and discusses their limitations. Bartering exchange ring based incentive mechanism was found to have the potential of fulfilling all the proposed design requirements. It organizes the nodes with asymmetric interests in the bartering exchange rings, enforcing the nodes to contribute while consuming. However the existing bartering exchanging formation approaches only rely on historical search records which may lead to a risk of using out of date information. Moreover, these incentive mechanisms lack of accountability so that the self-interested rational nodes can still obtain complete resources and only contribute before finishing the consumption. A novel cluster based incentive mechanism (CBIM) is proposed in this thesis which enables dynamic ring formation by modifying the Query Protocol of the underlyingP2P systems. The query messages become the media that the nodes can use to publish supply and demand information on. The nodes can then cooperate to form a cluster through the query messages while searching. A cluster can be formed when every node publishes same number of requests and provisions in a query message and all the requests can be satisfied. Graph theoretically, a cluster consists of one or more bartering exchange rings. The CBIM also uses a reputation system to alleviate the effect of malicious behaviours. The nodes try to identify free riders by fully utilizing their local transaction information. The identified free riding nodes are blacklisted and thus isolated. The simulation results indicate that by applying the CBIM, the overall request success rate of the network can be considerably improved since the rational nodes are forced to become more cooperative and the free riding behaviours can be identified to a certain extent.