Discrete-time stochastic models designed to encompass various sequential measurements of telephone traffic are derived from assumptions of classical queueing theory. The models are linear and their coefficients depend on the measurement cycle length normalized by the mean holding time. Proposed descriptions are tested against the real data gathered by the supervisor of the French toll network. Forecasting methods are then proposed and implemented in the simulator of a testbed network with adaptive routing. Simulation results confirm the benefits of advanced routing concepts and allow us to classify various information patterns from the standpoint of their usefulness for traffic control.