一个面向OLAP的多维层次聚簇存储模式  

An Efficient Multidimensional Hierarchical Clustering Storage Schema for OLAP

作　　者：袁霖[1] 邹恒明[1] 李战怀[2] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200240 [2]西北工业大学计算机学院计算机软件与理论系,西安710072

出　　处：《计算机科学》2007年第9期110-113,124,共5页Computer Science

基　　金：国家自然科学基金(60073055)资助

摘　　要：文献[2]针对ROLAP提出的多维层次聚簇存储模式(MHC),极大地提高了查询效率。然而与ROLAP相比,MOLAP往往具有更高的存储效率和查询效率。这让人自然地联想到:如果能构造一个集二者优点为一身的混合型OLAP系统,以实现MHC,也许能进一步提高系统性能。作为这一设想的探索性研究,本文利用ORDBMS的可扩展性实现了这一原形系统:多维数据按维层次分块聚簇,其中每个分块以数组ADT存储,分块间以B^+树索引聚簇。实验表明,本文提出的MHC实现能有效减少存储空间,进一步提高查询性能。ROLAP system can be accelerated by MHC （multidimensional hierarchical clustering） physical layout schema introduced by literature [2], while MOLAP system has advantages over ROLAP by high compression rate and efficiency. It is natural to ask if adapting the MHC idea for a mixed system that is a combination of MOLAP and ROLAP system can further improve the performance. As an initial step toward answering this question, we have implemented this kind of OLAP system that is based on ORDBMS. In our system, the data cube is chunked and clustered in terms of dimensional hierarchies. These chunks are stored as array-ADT and clustered with a B^＋ -Tree index. The experiments show that our MHC schema can save more storage space and further improve the performance.

关 键 词：联机分析处理 层次 聚簇 对象关系数据库 

分 类 号：TP311.131[自动化与计算机技术—计算机软件与理论]