摘要
尽管利用预计算可以提高OLAP的查询效率,但是,由于存储空间的限制,预计算整个数据方体是不现实的。最近提出的综合数据方体通过将数据单元进行等价划分的方法解决了这个问题。然而,当数据源发生改变的时候,要对这样的数据方体进行维护是很困难的,即使只有一条元组发生了变化,所有的聚集值都必须重新计算,代价非常高。实际上,在有些应用环境中,人们更关注查询响应的速度,在查询结果的精度上可以放低一些要求。本文提出了如何对近似的综合数据方体进行增量维护的方法。实验证明,这些方法是非常有效的。
It is often not feasible to compute a complete data cube due to the storage requirement. Recently proposed quotient cube addresses this issue through a partitioning method that groups cube cells into equivalence partitions. However, when the data source is updated, the aggregate values need to be recomputed even after one tuple is inserted or deleted. To keep the aggregate values to be always exact can prohibitively expensive in terms of time and/or storage space in a data warehouse environment. In many applications, it is sufficient to generate fast,approximate instead of full precise answers to queries. In this paper, we propose and examine techniques at the maintenance of an approximate quotient cube. Efficient algorithms are proposed and their effectiveness at storage and maintenance is investigated. A systematic performance study is conducted on different kind of data sets, which demonstrates our algorithms are efficient and scalable over large databases.
出处
《计算机科学》
CSCD
北大核心
2005年第9期100-102,共3页
Computer Science
