A connectionist system for ATM signalling analysis

Abstract

Signalling is an important index to show the system status of ATM systems. However, in the traditional network management strategies, the signallings are recorded first, and then analyzed later. This will cause serious and disastrous problems in the high-speed real-time networks like ATM. How to reliably analyze in real time the signalling status and quickly react is becoming an important issue in ATM networks. Therefore, in this paper, a connectionist system, based on neural network and expert system technologies, is designed for analyzing the signalling in ATM networks. The proposed system is divided into two modules: detection and diagnosis. The former module is responsible for detecting whether the information element of network signalling is normal or not and is built by a counter propagation neural network. When the detection module finds out the abnormal status, the diagnosis module is driven to infer the detailed reasons by counter propagation neural network and the expert system will execute the corresponding diagnostic actions. The developed system will be realized by VLSI technology and embedded in the control plane of broadband integrated services digital networks (B-ISDN). It's our belief that the designed chipset is a topnotch component for real-time error recovery in ATM networks.