Distribution Design in Distributed Databases Using Clustering to Solve Large Instances

In this paper we approach the solution of large instances of the distribution design problem. The traditional approaches do not consider that the size of the instances can significantly reduce the efficiency of the solution process, which only involves a model of the problem and a solution algorithm. We propose a new approach that incorporates multiple models and algorithms and mechanisms for instance compression, for increasing the scalability of the solution process. In order to validate the approach we tested it on a new model of the replicated version of the distribution design problem which incorporates generalized database objects, and a method for instance compression that uses clustering techniques. The experimental results, utilizing typical Internet usage loads, show that our approach permits to reduce at least 65% the computational resources needed for solving large instances, without significantly reducing the quality of its solution.