Fragment Allocation and Replication in Distributed Databases

We study the problem of designing a distributed database system. We develop optimization models for the problem that deals simultaneously with two major design issues, namely which fragments to replicate, and where to store those fragments and replicas. Given the difficulty of the problem, we propose a solution algorithm based on a new formulation of the problem in which every server is allocated a fragment combination from a set of combinations generated by a randomized greedy heuristic. The results of a computational study show that the algorithm outperforms a standard branch & bound technique for large instances of the problem.