Providing quality of service over time delay networks by efficient queue management

Abstract

The TCP/AQM dynamics is modeled as a time-delay congestion control system, in which the round trip time (RTT) has great influence on the system stability and network performance. Most of the existing active queue management (AQM) algorithms suffer from performance degradation, and fail to provide quality of service (QoS) when the RTT delay becomes remarkable. To address the problem, we present an efficient queue management method, which uses an enhanced Smith predictor to compensate for the RTT delay. An adaptation rule is designed for the gain of the estimated model via Lyapunov stability theory to improve the robustness of Smith predictor. Converting the Smith predictor to its equivalent structure, we derive the AQM controller from the internal model control technique and obtain an executable proportional-integral (PI) controller, which can be easily implemented due to its simple form. By simulations, it is demonstrated that our proposed method could improve QoS in large delay networks and outperforms other competitive AQM schemes in terms of stability and robustness.