Information theoretic analysis for queueing systems with quasi-random input

In this paper, information theoretic inference methodology for system modeling is applied to estimate the stationary distribution for the number of customers in single server queueing systems with service capacity utilized by a finite population. The customers demand i.i.d. service times. Three different models are considered. In Model I, a customer who finds the server busy can be queued, whereas in Models II and III, any customer finding the server busy upon arrival will make repeated attempts to enter service until he eventually finds the server free. Models II and III differ in the retrial policy. Numerical examples illustrate the accuracy of the proposed maximum entropy estimation when it is compared with the classical analysis.