A Markov chain model of a polling system with parameter regeneration (Q2475027)

The authors consider a polling system containing several stations and a single server switching (with non-zero random time) from a station to another station when it completes all jobs at a given station (exhaustive service). A particular case when only two stations are open receiving jobs at any time is studied, and the routing of server to visit stations follows an irreducible Markov chain. Jobs arrive at the open stations following (two) Poisson streams. The main assumption is that a (general) service time distribution and the arrival rates are regenerated at the start of each switching time using a regeneration measure depending upon the (ordered) pair of open stations only. The main goal of the paper is to establish stability conditions of the polling model. All results are based on constructive theory of Markov chains.The conditions for the existence of polynomial moments of hitting times for the related random walk are derived. The unusual phenomenon of a large null-recurrence region is obtained even for in the case of zero switching times. Instability conditions are also considered.