Date of Award

1990

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Sitharama S. Iyengar

Abstract

This thesis presents an efficient and integrated approach to integrity constraint checking for advanced database systems. The proposed approach essentially consists of three phases: constraint decomposition, global enforcement strategy and local enforcement tuning. The central theme of this dissertation is the development of constraint decomposition theory, which can be used to decompose each constraint formula into a set of constraint sub-formulas. The constraint sub-formulas derived from the decomposition satisfy the sufficient conditions imposed by the original constraint and are also much simpler and more efficient to check. The decomposition is done only once for each constraint, at the time of its definition. The global enforcement strategies, with the application of decomposition theory, are studied in both sequential and parallel environments. The physical characteristics of the database state is used to determine the checking order or to make assignment of constraint sub-formulas to processing elements. It is shown that the performance of constraint enforcement, with the application of the decomposition theory, is much better than that without the application of the decomposition theory. Local enforcement tuning methods are developed to further simplify each constraint sub-formula derived at constraint decomposition phase. Because of the simplicity of constraint sub-formulas, more efficient and simple methods can be used. Furthermore, more information, such as update types and update data, are available for the purpose of simplification. The fundamental assumption underlying our approach is that most database updates satisfy integrity constaints. A second assumption is that transactions are localized and database updates are nonuniform in their distribution. Based on these assumptions, the proposed approach will achieve a significant performance increase over previous approaches.

Pages

151

DOI

10.31390/gradschool_disstheses.4952

Share

COinS