P2P环境下数据管理系统上的Top-k查询.doc

P2P环境下数据管理系统上的Top-k查询 何盈捷 文继军 冯月利 王珊 （中国人民大学信息学院 北京 100872） 摘要 目前，大多数P2P系统只提供文件级的共享，共享粒度较粗，缺乏数据管理的能力。提出了一种在P2P环境下共享数据库的新框架：基于关键词查询的数据库共享。将每个节点上的数据库看成是一个文档集，用户不用考虑数据库的模式结构信息，简化了不同节点数据库模式间的映射过程，能够较好地适应P2P分散和动态的特性。把文档集上基于直方图的分层Top-k查询算法应用到P2P环境下数据库上的Top-k查询，将文档集和数据库的查询统一起来，一致对待。随着查询的进行，直方图可以自动更新，同时根据查询的结果对邻居节点进行调整，把那些真正包含Top-k结果的节点作为查询发起节点的邻居，具有自适应性。实验证明，基于关键词的数据库共享突破了传统的数据库共享模式，简化了数据访问方式，而基于直方图的Top-k查询算法提高了查询的效率。 关键词 Peer-to-peer，数据库共享，关键词查询，Top-k查询，直方图，邻居节点自调整 Top-k Query over Data Management System in P2P Network HE Ying-Jie WEN Ji-Jun FENG Yue-Li WANG Shan (Information School, Renmin University of China, Beijing, 100872) Abstract Most of existing peer-to-peer (P2P) systems only provide coarsely granular file-level sharing and lack of data management ability. A new framework of database sharing in P2P network is presented, which is based on keyword search over relational databases. In the framework, database on each peer can be seen as a text collection, users needn’t care about the schema information of the database, which greatly simplifies the schema mapping between different peers, and is more suitable for the decentralization and dynamicity of P2P network. Hierarchical top-k query processing algorithm based on histogram is applied in the data management system in P2P network to do top-k query, which unifies the query over text collections and relational databases. During the query processing, histogram can be updated automatically by using the returned top k results, and the neighborhood of the peer can be self reconfigured to let the nodes containing the real top k results to be its neighbors. Experiment results show that database sharing based on the keyword searching breaks through the traditional database sharing schema and simplifies the data accessing method, and top-k query based on histogram improves the search efficiency. Keywords Peer-to-peer, Database sharing, Keyword query, Top-k query, Histogram, Neighborhood self reconfiguration 引言 目前的