一种基于多重覆盖的结构化P2P搜索

本文提出了一种基于多重覆盖的结构化P2P搜索策略,mDOS(multi-Dimensional Overlays based Search).mDOS模拟社会网络中的小世界模型,根据结点的多重属性将结点组织成为多维树型结构.每一维上的任意两个结点根据其在树型结构中的位置形成语义距离.按照覆盖网络中实际距离与语义距离之间的一定函数关系构造多重覆盖网络.邻居结点和目标结点之间在网络中的实际距离可以通过他们的语义距离估计得到,形成基于结点属性的路由方向感.两个结点的有效距离取值多维中的最短语义距离.多重覆盖中的一个结点可以桥接在不同维上的邻居结点,缩小结点之间的平均有效距离,进而缩短搜索路径长度.mDOS还引入索引内容"懒惰维护"机制和路由表"被动修改"机制以克服结点动态性带来的覆盖网络维护代价.最后的实验表明mDOS可以降低搜索路径长度,降低维护代价,提高可用性.

A Multi-dimensional Overlays based Structured Peer-to-Peer Search

In this paper we propose a structured Peer-to-Peer Search,mDOS(multi-Dimensional Overlays based Search).mDOS simulates the Small World model in the social network,arranging the peers in multi-demensional tree structures according to peers' multi-attributes.The semantic distance between any two peers in any dimension is formed according to their positions in the tree.Multi-dimensional overlays can be constructed in therms of the relation between the virtual distance in the overlay and the semantic distance in tree structure.Then the virtual distance in overlay between the neighbors and the destination can be estimated according to the semantic distance,forming the attribute based sense of routing orientation.Because the effective distance between any two peers is the closest semantic distance among the distances in all dimensions,a peer can bridge the neighbors in different dimensions,shortening the average effective distance,and shortening the average search path length further.And the mechanisms of "lazily-maintained" and "passively-modified" are adopted in mDOS to deal with the overhead of overlay maintenance.The final simulation experiment shows that mDOS can improve the usability and facilitate the topology maintenance with the low search latancy.