A Greedy Approach to Query Processing for Different Query Distributions in Data Warehouses

Abstract

The Range-Based Index (RBI) reduces the response time for on-line decision support in the data ware-houses, especially for attributed with high cardinality in data records. However, partitioning the entire ranges into the number of bitmap vectors is a very critical issue for the RBI, since the task of fetching data from the disk for the attribute checking is very time-consuming. Moreover, from the history of users’ queries in the data warehouses, queries are frequently performed on data records with the same value or ranges of the attribute. It takes long disk I/O time when these data records are partitioned into different bitmap vectors. In this paper , we consider the history of users’ queries on the design of the partitioning strategy. Based on the greedy approach, we propose the GreedyExt and GreedyRange strategies for answering exact queries and range queries, respectively. These two strategies decide the set of queries to construct the bitmap vectors such that data records with the high frequency of the query can be quickly and directly accessed. Then, the response time can be reduced in most of situations.