【作者】 陈才；

【导师】 杨放春；

【作者基本信息】 北京邮电大学 ， 计算机科学与技术， 2010， 博士

【摘要】 由于对等网络开放、自由、自组织等特性导致的节点自私行为引发对等网络资源共享不足,在造成服务能力下降,破坏网络健壮性的同时,已经成为限制对等网络发展的重要因素之一。激励机制具有抑制节点自私性,鼓励节点协作、提升节点贡献的特性,是对等网络提供高效、可靠服务能力的有效保障。以往对等网络中激励机制的应用和研究往往需要中心服务器或者复杂体系结构的支持,造成较大网络开销。通过分析传统激励方法与抑制节点自私性方法的不足,基于对当前激励机制研究现状和趋势的分析,本文对对等网络中三个重要阶段—网络拓扑构建阶段、数据分发阶段、以及路由定位阶段中自私节点大量存在影响网络性能的重要问题进行了深入研究,提出无需复杂体系结构的激励方法。取得如下成果：1.提出了一种基于互惠贡献能力的网络拓扑调整优化方法。对等网络拓扑结构是实现资源高效搜索定位的基本保证。以往研究忽视了节点传输能力对网络拓扑调整的影响,导致优化调整后的网络拓扑尽管使合作节点得到汇聚,但资源分发性能依然低效。本文提出一种新的互惠贡献能力计算方法,该方法综合考虑节点自主行为特性以及传输能力,突出资源交易时间特性和交易连续性对互惠贡献能力的放大影响,达到更快识别自私、恶意节点的目的。在此基础上,提出了一个周期性、分布式、自调整拓扑优化算法。实验证明,本算法降低了自私节点对网络性能的影响,与现有网络拓扑调整方法相比较,有效缩短了资源查询定位跳数,提高了资源分发效率,并具有更小的网络开销。2.提出了基于节点贡献和贡献意愿的带宽资源分配方法。在系统带宽资源有限情况下,源节点合理选择带宽资源分配策略是实现数据内容高效传输的重要前提。以往的带宽分配研究或者没有考虑节点贡献进行随机等量分配,或者按照节点绝对贡献进行比例分配,忽视了节点异构能力在贡献中起到的作用。本文提出了一个两阶段带宽资源分配算法,实现了具有激励性质的公平带宽资源分配。首先,基于社会效用最大化原理,按照节点的贡献量进行比例分配。然后基于第一阶段分配结果,再针对节点贡献意愿进行带宽微调整。实验证明,整个算法在较少调整迭代次数内,实现了公平、高效的带宽分配。在基本不改变按节点绝对贡献分配资源的基础上,本算法小范围调整节点的带宽收益,提高了资源分配公平性,同时提升了对等网络的整体效用。3.提出了一种基于信念更新的节点消息转发合作策略。节点对系统消息的协作转发,是维护整个系统稳定运行的基础。针对对等网络路由定位阶段中,由于网络噪音导致节点产生合作性误判,从而难以形成有效合作的问题,本文将消息转发过程建模为重复博弈模型,提出了一种考虑网络噪音和误判情形下的基于信念更新的节点消息转发合作策略。节点根据自身的私有收益历史信息和不完美观测结果,利用贝叶斯规则更新自己对对手节点采取合作策略的信念,并独立做出是否合作的决策,最大化长期收益。这种合作策略在重复博弈中实现了贯序均衡,实验证明节点采取任何偏离这种合作转发的策略都不能获得额外收益,并且该方法确保合作节点获得比自私节点更高的平均收益。更多还原

【Abstract】 P2P systems become increasingly popular since its scalability with no central infrastructure, however, the inherited characteristics, such as open, free and self-organization etc, cause the selfish behavior of participant nodes. The selfishness and free-riding phenomenon have become a crucial problem and one of the important factors of limiting P2P network development. The incentive mechanism aimed to suppress the node selfishness and encourage the cooperation between nodes, is proposed by researchers to promote the node to contribute, and guarantee the effectiveness of P2P network offering high-efficient, reliable service ability. However, traditional incentive studying often needs a central server or the complicated system infrastructure in the past, leading to large expenses and cost.Based on current status and tendency of incentive mechanism research, we address several critical problems in distributed Peer-to-Peer environment, especially in three important stages:the topology construction and maintenance stage; content distribution stage and routing and message transmission stage. The contribution of this dissertation includes:1. Propose a reciprocated contribution capacity based topology optimization approach. The topology structure of p2p network is a basic assurance of high-efficient performance for resource search. In literature, little attention has been paid on the selfishness effects of topology. In order to tackle the problems that selfish (malicious) nodes occupy the center of the P2P network resulting in inefficient network performance, a local historical transaction based Reciprocal Contribution Capacity (RCC) of peers and an effective topology optimization approach (RCTO) is proposed. It is desirable for RCTO to establish connections between peers with higher RCC. Otherwise, selfish nodes would be recognized and forced to the marginal of P2P network. It is shown through experiments that RCTO can greatly shorten the search hops for authentic resource. Compared with counterpart RC-ATP, RCTO achieves higher resource distribution efficiency and incurs less cost.2. Propose a contribution willingness-based two phase bandwidth resource allocation approach (Cowtra). Previous studies addressing the free-rider problems focus on how to utilize peer absolute contribution to differentiate services quality, while neglecting the role of relative contribution of peers, leading to unfairness of resource (e.g bandwidth) allocation. At first stage of Cowtra, our approach guarantees the bandwidth of a source peer is distributed properly according to the peer contribution using social warfare maximum mechanism. Then at second stage, an iterative bandwidth micro adjustment process is conducted to softly reassign the amount of received bandwidth, with regard to the peer’s relative contribution, namely Willingness of Contribution (WoC). In this way, a rational peer would try to act more contributive and cooperative to get a better result. Cowtra promotes the fairness of bandwidth allocation, as peers get the resource in proportion with their contribution and WoC. Simulations demonstrate that Cowtra largely increases the fairness of bandwidth allocation and improve the peer utility in overall system.3. Propose a belief-based message relaying cooperative strategy. Known the cooperation between the nodes to relay messages unselfishly is the foundation that the whole p2p system operates steadily, we study the message-relay problem by modeling it as a repeated game with imperfect observation and under erroneous-prone judgment environment, and propose a strategy only relying on node’s own historical actions and its private imperfect observation of other node’s information. Then node updates the belief of the others cooperative action using Bayes’s rule, and take a proper action to maximization its long-term payoffs. Simulation and analysis demonstrate that selfish node who deviates from the proposed strategy can’t obtain any extra income, and this method guarantees that the cooperative node gets the average income higher than the selfish node.更多还原