Product-form stationary distribution of states for a class of queueing networks (Q1116557)

The author considers product form equilibrium probabilities for Markov queueing networks with multiple classes of customers. He introduces a weak reversibility, which he claims to be necessary and sufficient for a product form result. It is supposed that the k'th node receives traffic at the same rate of \(\lambda\) \(\epsilon_ k\) for each customer class; the \(\epsilon_ k\) appear to be arbitrarily specified, instead of arising out of a legitimate traffic equation. No justification is offered for such assumptions, which moreover enter implicitly into the definition of weak reversibility. The proof of product form rests on splitting each global balance equation into several local balance equations (again using the \(\epsilon_ k)\), and showing that the local equations are satisfied by a product form solution. Unfortunately, the author produces no argument to demonstrate the alleged validity of the local equilibrium equations.