SEQUENTIAL BOTTLENECK DECOMPOSITION - AN APPROXIMATION METHOD FOR GENERALIZED JACKSON NETWORKS

In heavy traffic analysis of open queueing networks, processes of inte rest such as queue lengths and workload levels are generally approxima ted by a multidimensional reflected Brownian motion (RBM). Decompositi on approximations, on the other hand, typically analyze stations in th e network separately, treating each as a single queue with adjusted in terarrival time distribution. We present a hybrid method for analyzing generalized Jackson networks that employs both decomposition approxim ation and heavy traffic theory: Stations in the network are partitione d into groups of ''bottleneck subnetworks'' that may have more than on e station; the subnetworks then are analyzed ''sequentially'' with hea vy traffic theory. Using the numerical method of J. G. Dai and J. M. H arrison for computing the stationary distribution of multidimensional RBMs, we compare the performance of this technique to other methods of approximation via some simulation studies. Our results suggest that t his hybrid method generally performs better than other approximation t echniques, including W. Whitt's QNA and J. M. Harrison and V. Nguyen's QNET.