基于节点簇的P2P随机漫步搜索

以Gnutella为代表的P2P系统通常会呈现复杂的网络结构,为此,文中提出了一种基于节点簇的随机漫步搜索算法.该算法利用节点簇来存储系统中文件的索引,通过将搜索过程限制于节点簇内部来提高搜索性能.基于数学模型的理论分析,文中给出了搜索性能上下界的数学描述.实验结果表明:搜索性能与簇的阈值c密切相关;c的建议值为系统中节点最大度值的一半,与普通随机漫步相比,此时稀有文件的搜索效率至少可以提高250%,文件索引的传输和存储代价可以减少一个数量级;该算法具有索引存储代价非常低、搜索效率高、易于实现和部署的优点.

Node Cluster-Based Random Walk Search in Peer-to-Peer Network

Most of the unstructured peer-to-peer applications exhibit a characteristic of complex networks. We are motivated by this fact and propose a novel cluster-based search strategy. The cluster is used for storing the file indices. The search process is constrained between the nodes at the cluster to improve the search performance. The analysis results based on mathematic model give the formulas of search performance, and experiment results certificate the search performance is closely related to the cluster threshold c. The search strategy improves the search effectiveness by 100% at least and decreases the transferring and storing cost by roughly an order of magnitude. The search strategy is stable, and can be realized and deployed easily with no need for global topology knowledge.