Determining the Relative Efficiencies of Computation Systems: A Linear Programming Approach

This paper proposes the application of the Data Envelopment Analysis (DEA) method - an application of fractional linear programming - to the measurement of the relative efficiency for a set of computing systems. DEA is based on the Pareto optimality of inputs and outputs. A computation system is said to be Pareto efficient if no other system can produce the same output with less of some resource (input) and not more of any other resource. 

By using DEA, we know how the overall efficiency of a computation system breaks down into technical and managerial components and how an efficiency rating (index) is obtained. Then, we identify inefficiency utilization of inputs (slack resources) and determine desired output levels for inefficient systems. Given this information, a managerial plan can be developed to reallocate resources and direct system tuning efforts. 

DEA can then be applied repeatedly to the set of system to determined the success of adjustments. DEA is applied to hypothetical data. A discussion of the results and suggestion for policies to govern resource balancing strategies concludes the paper. The derivation of the linear programming models in given an Appendix.