Abstract:
This paper considers a network consisting of $N$ nodes having $rN$ servers. At each node a Poisson flow of rate $\lambda(t)$ arrives. If a particle arrives at an empty node, it leaves the system. If there are servers at the node, then a server is chosen equiprobably, takes a particle, and passes it to a random node which is chosen equiprobably. The passing time has exponential distribution with mean one. The number of servers at each of $N$ nodes is bounded by $m$.
Keywords:Markov processes, nonlinear dynamical systems, global asymptotic stability, generating operator, convergence, mean field approximation, queueing theory.