Simulated Versus Real Life Data in Testing the Efficiency of Scheduling Algorithms

Abstract This paper compares and contrasts the use of simulated versus real data in testing the efficiency of scheduling algorithms. Five hypotheses, formulated to evaluate two important measures of algorithm efficiency, viz., CPU time and the number of iterations, are tested on over one hundred problems drawn from different sources. On the basis of these empirical results, it is shown that: (1) the real problems are easier to solve than the simulated ones; and (2) the Natural Order heuristic is more effective on real problems than simulated ones. Implications of these results for testing the efficiency of algorithms are discussed.