An Application of Neural Networks in the Connection Admission Control of ATM Networks

The connection admission control (CAC) in ATM networks is a flow controlling function that decides to allow or not a new connection joining in the network. This decision is usually based on the status of the current ATM network as its available resources, flow parameters and the quality of registered service (QoS) of new connections joining in the network as well as existing connections. This article proposes a CAC model in which neural network is used as a tool to maximize the number of admitted connections, the "profit" derived from accepted connections (thought from their QoS) or simply the used bandwidth.