Application of AR based model in proactive management of VBR traffic

Abstract

Variable bit rate (VBR) video is expected to be dominant in ATM based B-ISDN networks. The characteristics of VBR traffic such as high peak-to-mean bit rate ratio and stochastic variations of bit rate with respect to time, however, make bandwidth allocation quite difficult. Choosing peak rate as a criteria would mean under-utilization of the available network bandwidth, whereas, allocating bandwidth based on mean rate would lead to congestion in the network. The effect of large delay-bandwidth product, in ATM networks, also adds to the complexity and renders conventional reactive based congestion control schemes as highly inefficient. Bandwidth allocation based on the mean rate, aided with a proactive congestion control scheme that could predict the onset of congestion well in advance, is, therefore, the desired solution for managing VBR traffic in ATM networks. One such estimation and prediction based approach to traffic control and congestion avoidance is proposed herein. The traffic from the VBR source(s) is modelled as an AR modulated process. The call arrival is assumed to be a birth-death process. Using these models the current traffic state is estimated, the future traffic states are predicted and the occurrence of congestion in the network is forecasted. Some simulation results are presented to verify the validity of the model and illustrate the effectiveness of the technique.