Computing the conditional entry-state distribution in Erlang loss systems

Abstract

We consider an Erlang loss system with state-dependent arrival rates. Given the system is in steady-state and there are j customers being served, the system operator may wish to know about the distribution of arrival states for the j customers in service. Specifically, they may want the steady-state probability that any given customer entered when the system had i servers busy, given j customers are currently being served. We term this metric the conditional entry-state distribution and develop an algorithm to compute it.