Optimization problems and solution methods in the design of data distribution

Abstract The design of data distribution requires solving interrelated optimization problems, concerning data fragmentation and allocation, operations optimization and allocation, and the evaluation of system's performance for given data and operation allocations. Each problem can be solved with several different solution methods; thus, designing data distribution is a very difficult task. In this paper, we characterize each problem and the interactions between them, presenting a general framework for the design of data distribution. We present an entity-relationship schema of the design data dictionary, which stores all information useful during the design, and we use the data dictionary to specify the input and output data for each design problem. We then discuss how the design problems interact. Finally, we present an integrated toolset for the vertical partitioning of relations, which uses different solution methods, called DIVIDE and CONQUER. DIVIDE has a simple analytic model for performance evaluation, while CONQUER uses a detailed cost model. The complexity of methods and experimental results are presented.