Markov-modulated finite-source queueing models in evaluation of computer and communication systems (Q597000)

A single-server queueing system serving customers from \(N\) heterogeneous sources is considered. All the sources and the server have rates that depend on independent random environments, governed by ergodic finite-state Markov chains. A source remains inactive as long as one of its requests is in the system. Two asymptotic results are stated (proofs are only sketched): (1) If the arrival rates of all sources go to infinity in a certain way, the length of a busy period is asymptotically exponential. (2) If the service rates go to infinity in a certain way, the first time at which there are \(m\) customers waiting is asymptotically exponential.