基于浓缩数据立方的内存实化数据立方的构建  

Materialization of data cube in main memory based on condensed data cube

作　　者：陈长清[1] 程恳[2] 颜文跃[3] 

机构地区：[1]华中科技大学软件学院,湖北武汉430074 [2]华中科技大学继续教育学院,湖北武汉430074 [3]华中科技大学计算机科学与技术学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2008年第9期5-8,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：湖北省自然科学基金资助项目(2005ABA271);武汉大学软件工程国家重点实验室开放基金资助项目(SKLSE05-12)

摘　　要：为提高联机分析查询的速度,在浓缩数据立方的基础上,构建了元组级别的内存实化方法.以内存空间至少能容纳最细粒度数据小方为前提,在内存中构造两级Hash结构:第一级Hash结构存放最细粒度的数据小方以保证所有查询都可从内存中响应;第二级Hash结构按照聚集度高的小方元组优先、相同聚集度情况下尺寸小的小方中元组优先的选择策略,选择立方元组在内存实化.处理点查询时,首先从第二级结构中直接查找满足条件的立方元组.若对范围查询,则需从第一级结构中计算获得.由于最细粒度立方元组和其他一些粗粒度元组都在内存中,避免了费时的外存存取,数据立方更新和维护代价也得以降低.In order to improve the query speed of online analytical processing, a tuple-level materialization method for main memory is proposed based on condensed data cube. Under the precondition that there is enough main memory space to hold the finest granularity cuboid at least, two-level Hash structure is adopted in memory. The finest granularity cuboid was materialized as the first level to ensure all queries be answered from memory. The second level was constructed as following that the tuples with higher aggregate level were materialized in advance, and then tuples belonging to the smaller size cuboid first under the same aggregate level. For a point query, the tuple satisfying conditions will be searched from the second level first. For a range query, the result can be obtained by computing from the first level. Because the finest granularity tuples and other coarser granularity tuples are in main memory, so time-consuming accessing from disk is avoided, the update and maintenance cost is also reduced.

关 键 词：联机分析处理 浓缩数据立方 数据小方 元组 优化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]