Study And Implementation Of Network Congestion Control In High Speed Mobility Scenarios

With the rapid development of mobile networks and high-speed rails(HSRs),there is an increasing need for the quality of network services for people on HSRs.TCP is a reliable transmission protocol used by most service applications.In recent years,researchers have carried out large-scale measurement experiments on the performance of the TCP protocol on HSR networks,and found that compared to stationary or low-speed mobile state,the performance of TCP is severely degraded.However,existing TCP measurement studies in the high-speed mobile network mostly use TCP's default congestion control algorithm(CCA),without considering the impact of other new congestion algorithms on TCP's performance.Therefore,this research measures and analyzes the transmission performance of high-speed mobile networks from the perspective of the TCP protocol using different congestion control algorithms,and then designs an optimized congestion control algorithm.The main work and contributions of the thesis include two parts:1.Select the Beijing-Tianjin Intercity Railway to measure the transmission performance of three typical TCP congestion control algorithms including CUBIC,BBR and NewReno for high-speed mobile networks.It was found that handover is the main cause of the high-speed mobile network performance fluctuations.Transmission performance measurement results show that CUBIC algorithm achieves the highest throughput,BBR is slightly lower than CUBIC,and NewReno is the lowest;but CUBIC's RTT is more than twice that of BBR and NewReno.Based on the existing actual measurement data and theoretical analysis,this thesis constructs the network performance change model during high-speed movement.2.Design the HSRCC algorithm that is suitable for high-speed mobile networks.Based on the link detection,HSRCC estimates link bandwidth and transmission delay to set congestion windows.Meanwhile,HSRCC dynamically adjusts the transmission rate to detect more available bandwidth and determines the handover event in real time to quickly recover the congestion window.Simulation and real network experiment results show that the HSRCC algorithm can make better use of the limited link resources and achieve higher throughput than the typical algorithms described above.In the real HSR network,HSRCC improves the throughput of BBR by up to 20.6%and 13.9%higher than that of CUBIC,meanwhile the average RTT can still keep below 200ms.In addition,the HSRCC algorithm is still applicable in other common network scenarios.