Congestion Control In ATM Networks Based On Time-Delay Model

Future computer networks will become a high speed packed switched network that will be able to provide various service so as to support a wide range of network applications. ATM is just one of the key technologies that support the Integrated Service Digital Network (ISDN), which is adopted as a typical technique by the International Telecommunication Union (ITU) and is spread. But congestion of the traffic and the loss of data are main reasons that affect the quality of service. Design of an effective congestion control scheme is critical to the enhancements of the network performance. The paper is based on it and the control theory is introduced to solve the congestion of controllable flow.The paper begins with the foundation of ATM network, and introduces the basic work principle of networks and the feedback mechanism. Then, the normal data networks model is given and ABR feedback control model is set up based on per-flow buffer model. The feedforward disturbance compensation system based on pure Smith congestion control is designed, which regards the ABR bandwidth as the system disturbance. The queue length can be controlled beneath the threshold value and the link utilization is improved.Propagation delays could have an adverse impact on the stability and performance of congestion control algorithm in high-speed ATM networks. Smith principle is a good scheme for overcoming pure time-delay,but it is very sensitive to model estimation error. Therefore the pure Smith controller can't stable the networks. Three robust Smith congestion controllers based on per-flow buffer model are designed. By regulating the parameters, the QoS (quality of service) of ABR flow could be improved on the condition that there exits uncontrolled traffic flow in the bottlenecked switch, which can guarantee the stability of the networks. But the networks topology model that multiple VCs (virtue circuits) sharing the same bottlenecked switch is popular. Two kinds of robust Smith congestion controllers are proposed under this assumed networks model.Employing the small gain theory can solve the problem of robust stability. Then, when the RTT (Round-trip time) of every VC fluctuates around the estimated value, the steady value of queue could track the specified expected value and the oscillation is very small.In the final, the structure of networks topology is assumed more complex: the RTT and the numbers of VC is changed with time. Based on this assumed networks, a simple proportion congestion controller is designed. At last, by proving the theorem and the simulations, the proposed congestion controllers can guarantee the "piece-wise" stable and the QoS.