Poisson hypothesis: combinatorial aspect (Q2432949)

Assume that we have a large network of servers through which customers are traveling, being served at different nodes of the network. If the node is busy, the customers wait in a queue. Customers enter the system via some nodes, and customer flows into the system from the outside are Poissonian. Service time at each node is random, with some fixed distribution that depends on the node. The paper investigates the stationary distribution \(\pi _N \) at a given node \(N\). The recipe of the Poisson hypothesis for the approximate computation of \(\pi _N \) is as follows: Consider the total flow \(F\) of customers to the node \(N\). Replace \(F\) with a constant-rate Poisson flow \(P\), the rate being equal to the average rate of \(F\). Compute the stationary distribution \(\widehat \pi _N \) at \(N\) corresponding to the input flow \(P\). The claim is that \(\widehat \pi _N \approx \pi _N \). The paper gives an approach to the Poisson hypothesis. It uses a purely combinatorial statement concerning certain problem of placement of nodes on the real line.