Authors:
Yaokai Feng
;
Akifumi Makinouchi
and
Hiroshi Ryu
Affiliation:
Graduate School of Information Science and Electrical Engineering, Kyushu University, Japan
Keyword(s):
OLAP, multidimensional index, R*-tree, range query
Related
Ontology
Subjects/Areas/Topics:
Data Warehouses and OLAP
;
Databases and Information Systems Integration
;
Enterprise Information Systems
Abstract:
Multidimensional indices are efficient to improve the query performance on OLAP data. As one multi-dimensional index structure,R*-tree is popular and successful, which is a member of the famous R-tree family. We enhance the R*-tree to improve the performance of range queries on OLAP data.First, the following observations are presented. (1) The clustering pattern of the tuples (of the OLAP data) among the R*-tree leaf nodes is a decisive factor on range search performance and it is controllable. (2) There often exist many slender nodes when the R*-tree is used to index OLAP data, which causes some problems both with the R*-tree construction and with queries. And then,we propose an approach to control the clustering pattern of tuples and propose an approach to solve the problems of slender nodes, where slender nodes refer to those having a very narrow side (even the side length is zero) in some dimension. Our proposals are examined by experiments using synthetic data and TPC-H benchmar
k data.
(More)