Study On End-to-End Congestion Control And Quality Of Service Of Networks

Congestion control and quality of service are two interrelated research subjects. Due to the end-to-end essence of TCP, the source algorithm of congestion control can't acquire enough information of intermediate network to control traffic ideally. In order to solve this problem, the link algorithm of congestion control that can make use of traffic information of network has been one of the research focuses in recent years, especially active queue management mechanisms. On the other hand, the current Internet is based on the best-effort model, which can't provide any QoS guarantees. However, rapidly emerging multimedia applications require certain QoS guarantee. So QoS guarantee is indispensable to future networks, especially in wireless networks, the limited bandwidth resources make it a challenge problem. Focusing on these questions, the dissertation makes some research on traditional fixed network, cellular network, Ad Hoc network from network layer, transport layer, MAC sub-layer respectively.A method based on the change rate of average queue length to enhance RED is proposed. RED can greatly alleviate the congestion status of networks. It allows networks to achieve high throughput and low average queuing delay at the same time. However, the resulting average queue length is quite sensitive to the level of congestion and to the parameter settings of RED. To alleviate the problem of parameter dependence of RED, a new adaptive method based on the change rate of the average queue length to tune maxp was proposed. Extensive simulations show that the proposed mechanism can improve the robust property of RED considerably and outperforms the existing methods.A rate-based method to enhance BLUE is proposed. A significant change in the number of connections can invalidate the parameter settings of BLUE and cause the queue backlog oscillation between loss and under-utilization. Analysis and experiments show that BLUE is not strong enough to adapting to dynamic network environment. To solve this problem, an improved method is proposed though adopting the adaptive thought. The main idea of it is adaptively adjust the marking probability of BLUE according to the arriving rate of packets. Extensive simulation experiments show that the proposed method can enhance the robustness of BLUE considerably under dynamic network environment. Compared with BLUE, it can improve link utilization and decrease packet loss rate at the same time.An adaptive method to guarantee bandwidth fairness among TCP flows. The end-to-end TCP congestion control based on the window mechanism cause the throughput of TCP connection inversely proportional to the connection Round Trip Time (RTT). To mitigate TCP bias to short RTT connections, a differentiated service traffic conditioner can ensure connections with long RTT do not starve when connections with short RTT get all extra resource after achieving the target rate. The existing methods for RTT-aware conditioner work well when the congestion level is low. Otherwise, they often over-protect the flows with long RTT and starve short RTT flows. Due to this problem, an improved method based on adaptive thoughts is proposed and realized combining with the active queue mechanism BLUE. Extensive simulation experiments show that the posed mechanism is robust and outperform the existing methods.Strategies to improve TCP in wireless environments. Various kinds of methods proposed toimprove the performance of TCP in wireless environments are investigated. Most of existing methods are classified, and then the characteristic, the advantage and disadvantage of them are evaluated. The various methods are compared on the aspect of throughput, fairness guarantee, and practicality. Based on these works, the basic strategic to improve the performance the TCP in wireless environments are proposed.Study on service differentiation of mobile Ad Hoc networks. A network model is proposed to support service differentiation for mobile Ad Hoc networks by combining a fully distributed admission control approach...