一种并行处理多维连接和聚集操作的有效方法  

An Efficient Method for Parallel Multi Dimensional Join and Aggregation

作　　者：薛永生[1] 黄震华[1] 段江娇[1] 张延松[1] 吕晓华[2] 

机构地区：[1]厦门大学计算机科学系,厦门361005 [2]浙江理工大学信息电子学院,杭州310018

出　　处：《计算机研究与发展》2004年第10期1661-1669,共9页Journal of Computer Research and Development

基　　金：福建省自然科学基金项目 (A0 3 10 0 0 8);福建省高新技术研究开放计划重点项目 ( 2 0 0 3H0 43 )

摘　　要：随着并行计算算法的完善和廉价、功能强大的多处理机系统的成熟 ,使得采用多处理机系统来并行处理多维数据仓库的连接和聚集操作成为当前有效提高OLAP查询处理性能的首选技术 为此 ,提出一种降低连接和聚集操作开销的并行算法PJAMDDC(paralleljoinandaggregationformulti dimensionaldatacube) 算法充分考虑了多维数据立方体的存储机制和多处理机分布系统的结构特点 ,在原有聚集计算多维数据立方体的搜索点阵逻辑结构的基础上 ,采用多维数据仓库的层次联合代理 (hierarchycombinedsurrogate)和对立方体的搜索点阵进行加权的方法 ,使得立方体数据在多个处理机间的分配达到最佳的状态 ,从而在分割多维数据的同时 ,提高了并行处理多维连接和聚集操作的效率 算法实验评估表明 。Along with the perfection of parallel computing algorithms and the maturity of cheap but powerful multiprocessing systems, parallel multi dimensional join and aggregation on data warehouse with the multiprocessing system becomes the preferred technology for efficiently improving OLAP operation In this paper, a new method(PJAMDDC algorithm) is proposed for processing time consuming join and aggregate operation In this method, by taking into consideration the characteristics of the storage mechanism of multi dimensional data cube and the structural characteristic of multiprocessing system and adopting hierarchy combining surrogate/search lattice of data cube with weight on the basis of the former logic structure of search lattice of data cube It advances the join and aggregate efficiency by achieving the optimal state for distribution of data cube between processors As illustrated by experimental result of the performance analysis, PJAMDDC algorithm is efficient when it is used for parallel multi dimensional join and aggregation on data warehouse

关 键 词：数据仓库 OLAP 层次联合代理 并行聚集查询 数据立方体 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]