An optimal scheduling of cells in ATM switch using neural network

Abstract

In asynchronous transfer mode (ATM), the traffic cells from various information sources are statistically multiplexed at the physical layer to efficiently utilize the network resources. An ATM switching node must successfully route the many arriving traffic cells to the correct output without collisions of cells in the switch fabrics while sustaining the user-required quality of services (QoS) for all application. For scheduling of cells in a switch, neural networks are noted for their ability to process large amounts of data quickly using a copious number of highly interconnected processors. In this paper, we propose an optimal cell scheduling algorithm for ATM switch using Hopfield neural network. The proposed algorithm finds a set of nonblocking cells with the ideal energy functions for Hopfield neural networks and efficiently minimizes the cell-delay time using a delay matrix scheme. Every cell transmission time, the algorithm gives the optimal scheduling which minimizes cell-loss and cell-delay time in buffer, and also solves the cell blocking and cell-sequence problems.