Queueing networks with path-dependent arrival processes (Q6089004)

This paper is devoted to a Gaussian model for an open network of queues having a path-dependent net-input process with evolution that depends on its early history. Continuous Gaussian queueing models typically arise as the limits of standard queueing models as the scale increases. The authors focus on a Gaussian queueing model that is the heavy-traffic limit of a standard queueing model with an arrival process that is a path-dependent stationary point process. More precisely, this Gaussian model arises as the heavy-traffic limit for a sequence of open queueing networks, each with a multivariate generalization of a Polya arrival process. Dependence between the external arrival processes to the different queues in the network is allowed, and is modeled with the help of superpositions of independent path-dependent processes. It is proved that the workload and queue-length processes then converge to heavy-traffic limits that are reflected Gaussian processes. For a stationary point process, path dependence is characterized by satisfying a non-ergodic law of large numbers, and the authors emphasize that the net-input and queue-length processes for their Gaussian model satisfy non-ergodic laws of large numbers with tractable distributions.