Eecient Access Methods for Multidimensional Data

Interest in multidimensional data has been steadily rising in the database community. Storage and subset retrieval of multi-dimensional data is necessary for scienti c, geographic, temporal and business data. Our research has concentrated on developing techniques for e cient loading, retrieval, and update of multi-dimensional data. Current commercial systems (with the exception of one) do not support multidimensional e ciently. Instead, they rely on traditional one-dimensional techniques, resulting in suboptimal access methods with poor performance for large databases. The one system that does provide multi-dimensional indexing only two and three dimensional indexing using simpler and less e cient algorithms than those developed in this work. Our prototype code has been handed over to MetaComp for integration into their intranet toolkit. 1 Background and Problem Description Interest in multidimensional data has been steadily rising in the database community. Storage and retrieval of multidimensional data is necessary for many business, scienti c, geographic, and engineering applications. Traditional secondary storage techniques such as B-trees and hash tables are not suitable for dealing with multidimensional data: by concentrating on one-dimension at a time, large quantities of extraneous data may be retrieved. A number of techniques designed speci cally for multidimensional data have been proposed [1, 4, 5, 9, 10, 12]. In the area of spatial databases industry lags behind as only one commercial system (Informix) incorporates true multidimensional indexing. Even that product only provides a small subset of the methods we have developed in the last two years. Several third party software products complement relational On-Line Transaction Processing Systems with multidimensional On-Line Analytical Processing (OLAP) capabilities. Since commercial relational systems do not support e cient multidimensional indexing, OLAP products create and maintain analytical databases which duplicate needed data from the transaction databases. These OLAP capabilities are restricted to business data, and require costly data replication algorithms due to the lack of multidimensional indexing support. The Illustra product implements only a subset of the e cient multidimensional database indexing and spatial data support capabilities that the PIs intend to focus on. We have developed e cient techniques for multi-dimensional indexing (as outlined in section 2) and delivered our prototype to MetaComp for incorporation into their Intranet Toolkit. An example application would be a business data base. Finding all employees between 30 and 35 years of age and earning more than $60,000 is an example of two dimensional region query. Adding additional dimensions such as zip code range would further reduce the number of qualifying records. XXX put in gure here.