The \(M/G/1\) queue with processor sharing and its relation to a feedback queue (Q1177917)

A single channel queueing system with infinite waiting room is considered. Input flow is Poisson, service times have exponential distribution. After service a customer either leaves the system or feeds back. The feedback probabilities depend on the number of services the customer has already obtained. Successive service times of a customer, feedback mechanism, input flow are independent. Let a tagged customer complete its \(k\)-th service. Consider times required for the successive passes through the system and the numbers of customers in the system at the service completions of the tagged customer. The main result gives the joint distribution of these random variables in steady state. The result is based on PASTA property and the fact that the joint process of successive service completion epochs and queue length vector at these service completion epochs is a Markov renewal process. If the feedback probabilities tend to one and the mean service time tends to zero in such a way that the mean total required service time remains constant, then the system under consideration can be thought of as an \(M/G/1\) processor sharing queue. This allows to analyze the sojourn time distribution in the \(M/G/1\) processor sharing queue. A more general processor sharing queue where a finite number of customers are always present is considered.