A NOVEL NEURAL-NETWORK TRAFFIC ENFORCEMENT MECHANISM FOR ATM NETWORKS

The Asynchronous Transfer Mode (ATM) principle has been recommended by the CCITT as the transport vehicle for the future Broadband ISDN netw orks. In ATM-based networks, a set of user declared parameters that de scribes the traffic characteristics, is required for the connection ac ceptance control (CAC) and traffic enforcement (policing) mechanisms. At the call set-up phase, the CAC algorithm uses those parameters to m ake a call acceptance decision. During the call progress, the policing mechanism uses the same parameters to control the user's traffic with in its declared values in order to protect the network's resources and avoid possible congestion problems. In this paper, a novel policing m echanism using neural networks (NN's) is presented. The mechanism is b ased upon an accurate estimation of the probability density function ( pdf) of the traffic via its count process and implemented using NN's. The pdf-based policing is made possible only by NN's. This is due to t he fact that pdf policing requires complex calculations, in real-time, at very high speeds which is not feasible via conventional mathematic al approaches. The architecture of the policing mechanism is composed of two inter-connected NN's. The first one is trained to learn the pdf of an ''ideal non-violating'' traffic, whereas the second is trained to capture the ''actual'' characteristics of the ''actual'' offered tr affic during the progress of the call. The output of both NN's (which is an accurate estimate of the traffic bit-rate fluctuations in the ne xt measurement period) is compared. Consequently, an error signal is g enerated whenever the pdf of the offered traffic violates its ''ideal' ' one. The error signal is then used to shape the traffic back to its original values. The reported results, prove that our policing mechani sm is very effective in detecting (and policing) atl possible kinds of traffic violations (e.g., peak and mean violations).