Estimation of View Size Using Sampling Techniques

Abstract

Online analytical systems based on multidimensional data models enable analysts and decision makers to have fast and interactive responses to queries posed against the data stored in very large databases. Materialized views can speed up the execution of many queries. Materialization of all the possible views is constrained by the storage space and maintenance cost and one needs to select a subset of views to be materialized. Algorithms to determine the views to be materialized need the size of views as input. However, often the task of determining the actual number of rows present in a view can itself be computationally very expensive because of the large sizes of the tables and views. Counting actual number of rows present in each view takes considerable time in an extremely large database environment. Thus, methods to determine a good estimate of the size of a view in reasonable time instead of determining its actual size are very important. We explore the use of sampling to estimate the size of views. In this paper, we propose a hybrid estimator that takes into account the degree of skew in the data and combines Jackknife estimator with the Schlosser’s estimator and guaranteed error estimator to estimate the size of the view more accurately. The proposed hybrid estimator has been used to estimate the view size and simulation results show better estimated values as compared to individual estimators.