Leaky bucket mechanism with learning algorithm for ATM traffic policing

The traffic policing in ATM networks is one of the most critical preventive congestion control mechanisms. It intends to ensure that each source conforms to its traffic parameters negotiated during the Call Admission Control phase. In this paper, we enhance the known Leaky Bucket mechanism with a Learning Algorithm in order to police the distribution of the traffic, observing the values that the counter of the Leaky Bucket takes. As it will be shown, tighter and faster control is achieved by the proposed methodology, for any type of traffic source, resulting in more statistical gain and better guarantee of the QoS constraints.