一种基于CUBE的语义OLAP的I/O优化算法

A Method of I/O Optimization Algorithm Based on Semantic OLAP of CUBE

数据仓库中多维数据立方计算对联机分析有着极为重要的作用,完全实例化的数据立方体上进行查询是非常简单的,只要返回所需数据即可,但是,实际中完全实体化一个数据立方体开销是非常大的,所以,既能缩减存储空间又能将数据立方体进行完全实体化针对查询来说是一个非常重要的研究问题。而完全实体化的数据cube是非常庞大的,那么在生成过程中I/O问题一直是影响性能的一个很重要的指标。有效的提高I/O操作,是我们解决问题的主要切入点。本文提出基于语义OLAP的Dwarf算法的频繁访问I/O问题提出了一个优化算法。实验表明,新的算法提高效率在10%以上。

The computation of multi-dimension cube in data warehouse is of much importance.Dwarf is a highly compressed structure for computing,storing data cubes which can be materialized completely.During the constructing process,each closed node is stored in disk,while the computing of unit ALL needs access the closed nodes in the disk frequently.For avoid accessing the unnecessary closed nodes in disk,in this paper,we propose an optimized algorithm for the frequent I/O operation.The algorithm in this paper is an optimized algorithm for the frequent disk I/O access in Q-Dwarf.The property of this algorithm guarantee that once the nodes were written to disk,they will not be read out again.the one-time write to disk,not requiring read out again.The new algorithm has shown an notably increasing performance in our experiments.