ABSTRACT
In this paper, the information theoretic principle of minimum relative entropy (MRE) is applied, as a method of inference [4, 5], to determine the finite buffer stationary queue length distribution {qld) given, as a prior estimate, the qld of the corresponding infinite buffer queue. Over recent years, this principle has been applied towards the analysis of finite queues and arbitrary queuing networks with or without server vacations with generalized exponential (GE) processes, which are completely described by their first two moments (cf., [6, 7]). At the single infinite GE-type queue level, the MRE solution is exact while at the GE-type network level, relative entropy minimization provides costeffective queue-by-queue decomposition algorithms (cf., [6, 7]). These earlier analytic results serve further to establish the MRE inference technique and motivate its application towards the analysis of a wider class of queuing systems and networks with more complex interarrival and service time processes.
