Sensitivity analysis of performance for M/G/1 queueing systems

Queueing systems which are not Markov-type are often used as models to study practical engineering problems, for example, communication networks. In this paper, the problems of sensitivity analysis of the steady-state performance for an M/G/1 queueing system are discussed by studying its embedded Markov chain. The sensitivity formulae of the steady-state performance are given by the potentials of the embedded Markov chain. Since the embedded Markov chain is much simpler than the semi-Markov process that is used to describe the system states, the results in this paper will be very convenient for simulating computation of performance sensitivity and optimization of the system.