Poisson traffic processes in pure jump Markov processes and generalized networks (Q1348616)

In the first chapter of this paper conditions are presented under which embedded point processes in a pure jump Markov process with general state space are Poisson processes. The results are essentially known but are proved here by means of the dual predictable projections of the point processes and the time-reversed process. Using these results a simple proof of an also already known PASTA type theorem is given. In a second part of the paper a network of symmetric queues is considered which has regular customers as well as negative ones (also known as signals). The duration between successive arrivals of negative customers are phase-type distributed. A sufficient condition is given for a suitably defined Markovian state process to have a product-form stationary distribution, for which an explicit expression is stated.