Study On Receiver-Centric Congestion Control In Wireless Networks

The wireless networks are becoming increasing prevalent in recent years. The performance of TCP rapidly degrades in wireless networks due to the impact of high bit error rate. Therefore, developing a congestion control mechanism that can be adaptive to the wireless character has become a hot topic. In this dissertation, we proposed a receiver centric mechanism to deal with this problem.Firstly, we proposed a performance evaluation model to analyze the impact of packet loss on the transport layer. Since the model combines the character of receiver -centric congestion control mechanism and wireless packet loss, it can analyze the effect of wireless packet loss to the transport layer; moreover, it can point out how to deal with wireless packet loss at the transport layer.Secondly, we studyed the end-to-end improved mechanism. We proposed a receiver -centric and rate based congestion control mechanism. In this mechanism, the receiver sets the sending rate based on both the estimated bandwidth and Additive Increase and Multiplicative Decrease (AIMD) mechanism. The sender sends data packets according to the sending rate set by the receiver. Theoretical analysis and simulation results show that our mechanism can improve the network throughput and fairness. Furthermore, we proposed a fuzzy logic congestion control mechanism to alleviate the effect of delay jitters and improve the congestion control accuracy. This mechanism takes the factor of network resource utilization, bandwidth variation trend and packet drop reasons into consideration. Simulation results show that the mechanism can effectively utilize the network resources and avoid congestion.Thirdly, we proposed a MAC layer congestion control mechanism. The MAC layer adjusts its sending rate based on the wireless packet loss and feedback this information to the transport layer. Thus the transport layer congestion control mechanism can only treat the network queuing congestion while free from the effect of wireless drops. Simulation results show that this mechanism can improve the network throughput and fairness. Furthermore, we proposed an AP delay REQ mechanism in WLANs to deal with the TCP unfairness problem. The AP dynamically suspends the transport layer REQs according to the congestion control information. Theoretical analysis, simulation results and experiments show that this mechanism can significantly improve the network fairness and throughput.To improve the friendliness between traditional TCP and our proposed mechanism, we proposed a receiver assisted congestion control mechanism. In this mechanism, a receiver sets and feeds the advertised window back to the sender according to the estimated bandwidth. Once the sender receives the advertised value, it employs the AIMD algorithm to calculate an appropriate size of the congestion window. Simulation and experiment results show that our mechanism can make better use of network bandwidth, mitigate the effect of wireless losses, therefore improve the network performance and be friendly with current TCP.