This paper presents the results of an empirical analysis of a previously proposed algorithm for buffer allocation in finite open acyclic general-service multi-server queuing networks. Based on the computational results, it is concluded that the processing time of the algorithm depends on the number of network servers (as expected) but is independent of the squared coefficient of variation of service time. It is also concluded that the obtained allocations are robust and that the approximations for the performance measures are accurate, as verified by simulation. Finally, it is found that it is not straightforward to develop heuristic rules to allocate multiple servers in the topology without applying a buffer allocation algorithm to determine the optimal configuration.
Optimization; Performance evaluation; Stochastic process; Design of experiments
