VLSI for ATM network management

The development of VLSI circuit modules that can be easily integrated to support traffic shaping and multiplexing at the link level by combining analog signal processing techniques with the principles of neural network learning is described. These modules are designed to work in a broadband asynchronous transfer mode (ATM) network in which the ratio of propagation delay to cell transmission time makes traditional end-to-end flow and congestion control techniques unsuitable. Circuit solutions for the characterization problem (the measurement of a statistical index such as the average interarrival time to characterize the short-term behavior of a traffic source) and the prediction problem are discussed.<
>