Maximum entropy two-station cyclic queues with multiple general servers (Q1097160)

The efficient implementation and extension of various approximate methods for general queueing networks require the study of two-station cyclic queues. In this paper the principle of maximum entropy is used to analyze two-station cyclic queues with a fixed number of jobs, multiple servers, general service times and first-come-first-served (FCFS) and/or processor-share (PS) and/or ample or infinite servers (AS) scheduling disciplines. New robust ``one-step'' recursions for the queue length distribution are derived and asymptotic connections to infinite capacity queues are established. Links with ``Birth-Death'' and more general global balance solutions are determined by making use of the Generalized Exponential (GE) model of the form \[ f(t)=(\frac{C^ 2-1}{C^ 2+1})u_ 0(t)\quad +\quad \frac{4\mu}{(C^ 2+1)^ 2}\exp (-\frac{2\mu t}{C^ 2+1}),\quad t\geq 0, \] where 1/\(\mu\) is the mean value and \(C^ 2\) the squared coefficient of variation, to represent general service- time distributions with known first two moments. Extensions to load dependent servers with Bernoulli feedback are also presented. Numerical examples provide useful information on how critically system behaviour is affected by the distributional form of service time and simple bounds for typical performance measures such as throughput and mean queue length are defined. Moreover, the utility of the work as a ``building block'' for the approximate analysis of a general central multiple server model is demonstrated.