In this paper augmented infinitesimal perturbation analysis (APA) is used for gradient estimation in a hybrid simulation/analytic system for computer communications network capacity planning called ATM Network Design (AND) System. Usage of APA provides an alternative to traditional network planning methods that assume independence of packet service times. A steepest descent algorithm determines locally optimal minimum average network delays based on these APA estimates. The algorithm employs an Armijo line search. In addition a linear, separable capacity constraint is assumed. It is assumed that input regulators (e.g., leaky bucket regulators) are used at the source of each virtual circuit. All virtual circuit external arrivals are modeled as independent Poisson processes. Regenerative simulation is used.

Two new analytic approximations are given for the mean delay in the timed token bus network specified in MAP (manufacturing automation protocol). Each station on the network has two kinds of timer for controlling the maximum amount of time for data transfer before giving up the token. High-priority stations would set large values in the token holding timer (THT) and transmit until the timer expires. Low-priority stations compare the elapsed time since the last token arrival (measured cycle time) to the token rotation timer (TRT) setting, and transmit only for the time left in the timer. The first approximation, based on Fuhrmann's bound for the limited-k service discipline, is for the symmetric case where all the stations have the same traffic load and the identical THT or TRT settings. The approximation has been compared to simulation results, and found to be quite accurate for a representation factory network, with THT or TRT set at two and five times the average total walk time. The second approximation is for the asymmetric case where there is one saturated low-priority (limited-TRT) station, and the rest of the stations in the network have high priority with limited-THT service. These two formulas will be useful in the planning and operation of MAP networks by providing guidance in the choice of timer settings for meeting different priority service requirements.>

Expressions are obtained for the mean excess token holding time for general service time distributions. They were used in the mean message delay approximations for both homogeneous token holding timer (THT) networks and nonhomogeneous THT and token rotating timer networks. Comparisons with simulation results indicated that these approximate formulas are quite accurate. The nonhomogeneous case considered here is only one of many possible mixes of low- and high-priority stations in a network. However, for the purpose of planning manufacturing automation protocol networks, only approximate and worst-case analyses are needed to determine the size of the network and the traffic characteristics for certain performance specifications. The formulas derived here can be used to provide guidance in the choice of timer settings for meeting various priority service requirements.>