Analysis of tradeoff between data accuracy and performance of databases

Abstract

Databases in large organizations are used for supporting the storage and manipulation of both operational and aggregate data. The operational data is characterized by frequent updates based on changes to values of items in the real world. Aggregate data is used in statistical analysis for decision making, forcasting and formulation of business strategies. The accuracy of the results of statistical processing depends on the processing resources allocated to the update transactions relative to the query and statistical reporting requests. Different types of statistical processing may require varying degress of data accuracy. For example, long range planning decisions may require less accurate data than those needed for tactical decisions. In this paper we propose a model for estimating the accuracy of a database and for the planning of computing resources based on database system performance and accuracy requirements of various groups of users.