Generalized end-to-end performance estimation of ATM networks using combined simulation model and neural network

Without the practical design method that guarantees every users' specified quality of service, ATM technology can not be the solution for building a public wide-area broadband integrated services network. In this paper, we propose a generalized and light-weight computational method to estimate the end-to-end performance of an ATM network. Degradation of quality of service at the terminating node is analyzed in view of the dynamics of the switching network and a neural network model for identifying the characteristics of outgoing traffic in a single switching node is suggested. The proposed ATM network design procedure to combine the simulation model and neural network model is easily adaptable and reusable.