Reducing bandwidth allocation delay in a DVB-RCS network using Bayesian Neural Network

Bandwidth on Demand requesting and assignment has became a vital requirement for the ever growing satellite networks with an expensive bandwidth. To reduce the signaling overhead, the bandwidth requests should be sent with long enough periods. But the problem arises when the input traffic rate of a terminal increases dramatically between two successive requests causing an increase in queuing delay. To take this into account, we used a Bayesian Neural Network for modeling the traffic. Then we can use this accurate model to predict the traffic entering terminals. Thus we can forecast the traffic bursts and send more efficient bandwidth requests.