Multidimensional indices are very helpful to improve query performance on multidimensional data including relational data in ROLAP systems. The existing multidimensional indices are directed to "queries with all dimensions" (called QAD in this study). That is, the dimensions used in each query are all the dimensions in the whole space. However, in many applications, especially in OLAP-related ones, the queries may be only with some (partial) dimensions (not all) of the whole space, which is called QPD (Queries with Partial Dimensions) in this study. This study focuses on range queries with partial dimensions (RQPD), which is popular in OLAP applications. If the existing multidimensional indices are used in RQPD, the dimensions unused in the query are thought as spanning the whole data ranges, which often lead to not-good search performance. In these cases, certainly, we also can construct many indices with all the necessary combination of dimensions. However, this is very space/time-consuming since many indices have to be constructed and some dimensions may be used many times in different indices, which is not always feasible. In this study, we propose a novel solution to RQPD problem. With our solution, only one index is necessary to such applications. The performance of our solution is discussed in detail and is examined by experiments.
|Number of pages||8|
|Journal||IPSJ SIG Notes|
|Publication status||Published - Jul 13 2004|