The use of learning algorithms in ATM networks call admission control problem: a methodology

Call Admission Control (CAC) is one of the most fundamental preventive cong estion control mechanisms in Asynchronous Transfer Mode (ATM) networks. Sev eral mathematical approaches that have been proposed in the literature, whi ch usually estimate the equivalent bandwidth that is required, achieve cons ervative approximations that result in reduced statistical gain and thus, i n under-utilisation of the network resources. In this study, a new methodol ogy is proposed which uses a Learning Automaton (LA) in combination with eq uivalent bandwidth approximations to reduce the percentage of overestimatio n. The learning algorithm that is used attempts to predict in real-time if a call-request should be accepted or not receiving as feedback a function o f an estimate of the equivalent bandwidth. As will be shown, the proposed m echanism, whose hardware implementation is feasible, exhibits remarkable st atistical gain compared with some classical CAC schemes of the literature a nd distinct improvement of the equivalent bandwidth approximations, Finally , some issues for extending this work are also discussed. (C) 2000 Elsevier Science B.V. All rights reserved.