Prototype Validation of the Trapezoidal Attribute Cardinality Map for Query Optimization in Database Systems.
Murali Thiyagarajah, B. John Oommen · International Conference on Enterprise Information Systems · 1999
Current database systems utilize histograms to approximate frequency distributions of attribute values of relations. These are used to efficiently estimate query result sizes and access plan costs and thus minimize the query response time for business and (non commercial) database systems. In two recent works (Oommen, Thiyagarajah 1999a, Oommen, Thiyagarajah 1999b) we proposed two new forms of histogramlike techniques called the Rectangular and Trapezoidal Attribute Cardinality Maps (ACM) respectively. Since these techniques are based on the philosophies of numerical integration, they provide much more accurate result size estimations than the traditional equi-width and equi-depth histograms currently being used by many commercial database systems. In (Oommen, Thiyagarajah 1999a; Oommen, Thiyagarajah 1999b) we also provided a fairly extensive mathematical analysis for their average and worst case errors for their frequency estimates—which, in turn, were verified for synthetic data. This paper reports the prototype validation for the Rectangular-ACM (R-ACM) for query optimization in real-world database systems. By using an extensive set of experiments using real-life data (U.S. Census 1997, NBA 1992), we demonstrate that the T-ACM scheme is much more accurate than the traditional histograms for query result size estimation. We anticipate that it could become an invaluable tool for query optimization in the future.