Efficiency of the Prediction of High Priority Traffic in Enhancing the Rate Based Control of Low Priority Traffic
Kimmo Pulakka, Jarmo Harju · 2000
In this paper, we study how effectively the rate based control of a low priority data transmission service in a packet switched backbone network can be implemented, when the control decisions are based on the predictions of the amount of high priority traffic. ANFIS (adaptive neuro-fuzzy inference systems) predictors are used for traffic prediction. In our service model, all control functions of the low priority data transmission service are distributed to the edge switches of the backbone network. The routes and data rates of the low priority data flows are iteratively controlled by the control system, according to the predicted data rate variations of the high priority data flows. The efficiency of the service model has been tested by simulations. The emphasis is on the effects of different traffic distributions of the high priority data flows and on the impact and importance of the amount of overhead caused by the traffic prediction and the transmission of control information.