A heuristic approximation for reducing problem size in network file allocation models

Abstract

Listen

Translate article

The design of distributed processing systems has increasingly drawn the attention of systems designers as more and more organizations recognize the advantages of decentralizing operations. However, determination of an optimal or preferred distributed data base configuration is a nontrivial task. A viable model for allocating files to nodes in a distributed system must consider the tradeoffs between costs and service and the unique characteristics of the system, including the planned redundancy inherent in the file organization. It also must be solvable within the current state of the art. This paper presents a mixed-integer linear programming model which allows assignment of replicated and/or partitioned files to the nodes of a distributed network while considering query, update and storage costs associated with the file assignments. The resulting model, though large, is solvable even for networks with a substantial number of nodes and links. Because the model relies on an average communication cost for determining file location, results are presented comparing the model performance to that of a model utilizing exact communication costs.