Distributed Congestion Control Based On Active Queue Management

With the rapid advances in the deployment of very high bandwidth links and the explosion of various types of network transactions, the need for efficient and high quality of transmission for diverse QoS requirements has been a hot research and development issue in computer networks. Therefore, a survivable replacement of congestion control mechanism used to facilitate the QoS implementation has been increasingly important.Traditional congestion control mechanism has deployed in the TCP protocol. The end user assumes the responsibility of congestion detection by perceiving packet loss event or monitoring the change of RTT. However, the user can not know the network state, which results in the blindness of TCP congestion response. Upon that, the distributed congestion control scheme based on active queue management (AQM) has introduced congestion control mechanism into intermediate routers.Firstly, a load-adaptive AQM algorithm named AOPC has been proposed in the dissertation, which eliminates the parameter tuning problem encountered in existing AQM algorithms. AOPC uses the measured packet loss ratio to configure the control parameters adaptively based on optimized second-order control system model. Theoretical analysis and experiments demonstrate that AOPC has enhanced the stability and responsiveness. Meanwhile, AOPC obtains high link utilization, low delay and jitter.Subsequently, the dissertation develops a stabilizing PID controller based on internal model control (IMC) principle, which eliminates the negative impact on stability caused by large communication delay. Simulation results show that the integrated performance of the proposed IMC-PID controller outperforms that of other AQM controllers as communication delays increase, and achieves high link utilization and small delay jitter.Finally, an explicit feedback congestion control protocol named ARROW TCP has been designed to improve the transmission efficiency, stability and convergence in multi-router network scenarios. Theoretical analysis and simulations illustrate that ARROW TCP obtains strong stability in multi-router networks and Web traffic scenarios. Meanwhile, it achieves max-min fair allocation and exponential convergence to efficiency and fairness within a constant time.The dissertation summarizes our research work in the last section. A further research direction on network resource allocation based on distributed congestion control scheme has been explored in detail afterward.