Key-Based Problem Decomposition for Relational Constraint Satisfaction Problems

Constraint satisfaction problems (CSP) are often posed over data residing in relational database systems, which serve as passive data-storage back ends. Several studies have demonstrated a number of important advantages to having database systems capable of natively modelling and solving CSPs. This paper studies the automated decomposition of the input CSP, borrowing another distinctive idea of relational databases, normalization, to improve its solution time. Experimental evaluations for two case studies show the potential benefit of the approach.