Time-scaling limits for Markov-modulated infinite-server queues
Stochastic Models , Volume 29 p. 112- 127
In this paper we study semi-Markov modulated M/M/$\infty$ queues, which are to be understood as infinite-server systems in which the Poisson input rate is modulated by a Markovian background process (where the times spent in each of its states are assumed deterministic), and the service times are exponential. Two specific scalings are considered, both in terms of transient and steady-state behavior. In the former the transition times of the background process are divided by $N$, and then $N$ is sent to $\infty$; a Poisson limit is obtained. In the latter both the transition times and the Poissonian input rates are scaled, but the background process is sped up more than the arrival process; here a central-limit type regime applies. The accuracy and convergence rate of the limiting results are demonstrated with numerical experiments.