Multi-objective Buffer Space Allocation with the Cross-entropy Method

The buffer allocation problem (BAP) has been widely studied by researchers while pursuing diverse research goals. Similarly, the cross-entropy method has been applied to a variety of optimisation problems with single objectives. In this article it is extended to the multiobjective case and proposed as a computationally economic approach to optimise at least two conflicting objectives of the BAP, namely throughput rate and allocated buffer space, while using computer simulation as evaluation function of small to large stochastic queuing networks of unreliable resources. No assumptions are made regarding the service time, timeto-failure and repair time distributions, and a general solution for obtaining the networkrelated Pareto front is proposed. The results for test networks indicate that reasonable Pareto fronts can be obtained via a low number of multi-objective solution evaluations using the modified cross-entropy method (CEM). (Received in April 2012, accepted in September 2012. This paper was with the author 1 month for 1 revision.)