基于XML中间件的分布式异构数据库的数据分片策略研究  

ON DATA PARTITIONING STRATEGY OF THE DISTRIBUTED HETEROGENEOUS DATABASE BASED ON XML MIDDLEWARE

作　　者：蒋勇[1,2] 谭怀亮[2] 李光文[1] 

机构地区：[1]湖南化工职业技术学院信息系,湖南株洲412004 [2]湖南大学计算机与通信学院,湖南株洲412004

出　　处：《计算机应用与软件》2009年第11期144-146,共3页Computer Applications and Software

摘　　要：主要研究XML中间件的分布式异构数据库的数据分片策略,以便能够并行处理XML查询。为了描述XML数据分片,提出了对象树、子树的概念,将一棵XML数据树分割成对象树和子树,对象树在各个站点上复制,子树标签化后均匀分片到各个站点上。为了实现用户数据分片查询工作量的均衡,对最大树深的子树采用减治法策略,减治法策略后的子树又分配到一个新的站点上。基于这一思想,提出并实现其算法。实验结果证明:在分布式异构数据库的检索、查询和访问中,基于对象树、子树的数据分片算法的性能要比PSPIB算法性能优越。This paper targets on studying data partitioning strategy of XML middleware based distributed heterogeneous database for the parallel process of XML queries. For describing the partition of XML data,the concept, objet tree and sub-trees, are presented in this paper. An XLM data tree can be partitioned into an object tree and the sub-trees. While the object tree is duplicated over all nodes,the sub-trees are labelled and evenly partitioned over all nodes. In order to realise a balanced workload of user＇ s data partitioning queries,the reduce-and-conquer strategy is carried on to the maximal deep-sub-tree, after that the sub-tree is then assigned to a new node. Based on the idea described above, in this paper it designs and implements this algorithm. The experimental results show that its speedup and scale up performances outperform the PSPIB algorithm in information retrieval, query and visit of the distributed heterogeneous database.

关 键 词：分布式异构数据库 文档对象模型文档 中间件 数据分片 对象树 算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP311[自动化与计算机技术—计算机科学与技术]