On the Efficiency of Resources Utilization in Strategic Peer-To-Peer Systems

This dissertation, "On the Efficiency of Resources Utilization in Strategic Peer-to-peer Systems" by Dingding, Guo, 郭丁丁, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Peer-to-peer (P2P) systems have achieved outstanding success during the past decades and much efforts has been put into investigating incentive mechanisms for strategic P2P networks. In the numerous studies of P2P system, efficient resource utilization has always been a critical factor in designing incentive schemes. Most existing studies try to encourage strategic peers to contribute more to the system, in order to maximize the resources. However, without considering how to effectively measure contributions and without adopting well-designed trading policies, simply motivating more contributions could lead to outcomes that do not match the original intention. This thesis, therefore, focuses on the investigation of efficient resource utilization in strategic P2P systems. First, it is found that increased contributions in terms of upload rate does not necessarily lead to better system performance. Observing that different chunks have different values to both the system and individual peers, a value-based metric is devised to measure contributions instead of using rate-based metrics. A variation of BitTorrent is also proposed, called value-based BitTorrent (VBT). VBT is found to effectively punish the strategic behaviors of an underreporting chunk map, and there is a positive correlation between investment and return for cooperative peers in VBT networks. Moreover, VBT always outperforms BitTorrent in terms of system performance. Second, taking the chunk value in the reciprocity process into consideration, the overpayment problem in a BitTorrent network is investigated, and four side effects of overpayment are identified. A new scheme is proposed to visualize overpayment and a series of metrics is proposed based on this method to quantify overpayment. The proposed value-based approach is found to be able to alleviate the degree of overpayment and consequently relieve the side effects of overpayment. Third, the performance of popular protocols in a P2P file-sharing system is investigated from the perspective of overpayment. These protocols are studied in two directions, when the measurement metrics are varied and when trading policy becomes tighter. The correlations between fairness, performance, and robustness are also examined. Finally, because overpayment is a fundamental problem of improper price setting, and auction is a widely used and effective method in setting prices in distributed systems. Auction is analytically proved that it is able to lead to optimal price without overpayment. However, most existing auction schemes are based on credit, which could induce a huge overhead in maintaining a monetary system, and monetary systems also have many inherent problems, such as inflation. The efficacy of applying an auction-like approach in P2P systems without money is investigated. A simplified version of the barter-based auction-like approach is tested in P2P file sharing and its overpayment degree is evaluated. Moreover, a novel barter-based auction-like approach is proposed for a P2P streaming system, and it is found that it can successfully punish strategic behaviors, with overall system performance outperforming a tit-for-tat strategy. DOI: 10.5353/th_b5223977 Subjects: Peer-to-peer architecture (Computer networks)