LB-SELA: rate-based access protocol for ATM networks

The characteristics of service independence and flexibility of ATM networks make the control problems of them very critical. One of the most fundamental is source policing. The most critical situation regards the control of bursty sources. Especially for this kind of traffic the estimation of the characteristics of the source is difficult. The known policing mechanisms cannot control effectively the negotiated traffic parameters. The authors enhance the “traditional” leaky bucket (LB) using a stochastic estimator learning algorithm (SELA) in order to “learn” the behaviour of the source and, if necessary, to react. This generalized mechanism improves dramatically the performance of LB and can be used for any bursty source. It achieves tighter and faster control and exhibits no computation overhead. The above features are investigated via simulation, and the extracted results demonstrate the effectiveness of this control