A Theory Refinement Approach to the Design of Distributed Object Oriented Databases

Distributed Design involves making decisions on the fragmentation and placement of data across the sites of a computer network. The first phase of the Distributed Design in a top-down approach is the fragmentation phase, which clusters in fragments the information accessed simultaneously by applications. Since Distributed Design is a very complex task in the context of the OO data model, we have presented in our previous works a strategy to assist distributed designers in the fragmentation phase of OO databases, which was divided in three phases: Analysis Phase, Vertical Fragmentation and Horizontal Fragmentation. The Analysis Phase defined the most adequate fragmentation technique (horizontal, vertical or mixed) to be applied in each class of the database schema, based on some proposed heuristics. Initial experiments using our proposed approach have resulted in fragmentation schemas offering a high degree of parallelism together with an important reduction of irrelevant data. In this work, we present a knowledge-based approach to the Design of Distributed Object Oriented Databases (DDOODB), through the use of Machine Learning. The objective of the work is to uncover some previously unknown issues to be considered in the distributed design process. In the present knowledge-based approach, we represent the Analysis Phase as a set of rules (the Initial Theory) that will be used as background knowledge to obtain a new set of rules (the Refined Theory) by applying the Inductive Logic Programming (ILP) technique, through the use of the FORTE system. This new set of rules will represent a Revised Analysis Phase that will propose good fragmentation schemas with improved performance. Although we address in this work the Analysis Phase for the class fragmentation problem in the DDOODB context, a possible future work is to apply the same theory refinement approach in other phases of our strategy, i.e., in the Horizontal and Vertical Fragmentation Phases, in order to improve their algorithms. Also, we may experiment with other data models, such as Relational Databases or Deductive Databases. Another idea of improving this work and show its applicability is to apply the resulting fragmentation schemas obtained from our Revised Analysis Phase to fragment databases used by very frequent applications such as Data Mining. This practice will help data miners in scaling-up their algorithms, considering that they usually deal with a great amount of data.