Non-stationary waiting times in a closed exponential tandem queue (Q1919818)

The paper considers a pair of single servers in series. Service times are exponentially distributed and the service discipline is FIFO at both the stations. Initially there are \(m\) \((n)\) customers in station 1 (2). No new customers enter the system. Using an imbedded random walk of the queueing process, the author shows that the distributions of (i) the time \(S_k\) spent in the system by the \(k\)th customer at station 1 and (ii) the time \(T\) needed to service all the \((m+n)\) customers, are expressible in terms of certain Erlang distributions. Thus only numerical integrations are required to obtain the densities of \(S_k\) and \(T\).