A state feedback control approach to stabilizing queues for ECN-enabled TCP connections

In this paper, we present an analytical TCP model that takes into account of several issues that were ignored in the other existing models (such as those in F. Kelly (2001) and V. Mishra et al. (2000)), i.e., the congestion window is not gradually decreased at the rate of /spl omega//sup 2/p/2, but suddenly halved upon receipt of congestion indication and the congestion window is halved at most once during one round-trip time (RTT). We also include the delayed ACK option in the model. With the use of state feedback control theory, we then design, based on the enhanced TCP model, an AQM controller to stabilize the queue length at routers. The performance of the new controller is shown, via ns-2 simulation to outperform several other schemes under a variety of network scenarios and traffic loads, in terms of fluctuation in the queue length, link utilization, and packet loss ratio.