A Framework for the Design of Distributed Databases

This work presents a framework to handle the class fragmentation problem during the design of distributed object databases. The framework works in the conceptual level, and thus uses the object data model to capture the application semantics represented by the user. The proposed framework integrates three modules. The heuristic module defines a set of heuristics to drive the fragmentation of object databases and incorporates them in a methodology that includes an analysis algorithm, horizontal and vertical class fragmentation algorithms. The theory revision module automatically improves the analysis algorithm through the use of an artificial intelligence technique named theory revision, using fragmentation schemas with previously known performance presented as examples. Finally, the branchand-bound module uses optimization techniques to perform an intelligent search for an optimal fragmentation schema through a larger space of hypotheses when compared to the space of hypotheses covered by the heuristic approach.