Towards self-learning adaptive scheduling for ATM networks
R.К. Mehra, Bharadwaj Ravichandran, J.B.D. Cabrera, D.N. Greve, Richard S. Sutton · 2002
We discuss ongoing efforts at Scientific Systems towards the development of effective strategies for traffic management of ATM networks using self-learning adaptive (SLA) techniques. We extended our previous SLA techniques to bursty traffic patterns and show how an approximation to SLA, called proportional feedback, can be used to manage real-time variable bit rate ATM traffic. Finally, we present results on the dynamic allocation of bandwidth in order to efficiently multiplex several variable rate MPEG-1 streams over a constant-rate link.