Some Numerical Experiments on Multi-criterion Tabu Programming for Finding Pareto-optimal Solutions

Decision making for complex systems is based on multi-criterion-optimization. A decision making support can be applied to find the Pareto solutions. Multi-criterion tabu programming is a new paradigm for that task. Similarly to rules applied in the genetic programming, tabu programming solves problems by using a tabu algorithm that modifies some computer programs. We consider the multi-criterion problem of task assignment, where both a workload of a bottleneck computer and the cost of system are minimized; in contrast, a reliability of the distributed system is maximized. Furthermore, there are constraints for the performance of the distributed systems and the probability that all tasks meet their deadlines. What is more, constraints related to memory limits and computer locations are imposed on the feasible task assignment. Finally, results of some numerical experiments have been presented. Key-Words: Tabu search algorithm, multi-criterion optimization, genetic programming