Study On The Measurement And Optimization Of Heterogeneous Network Transmission Protocol Performance

In recent years,with the continuous development of Internet technology and the great improvement of wireless networks,the number of mobile devices is increasing,and the increase in the size of users and the diversification of demand have put forward higher performance requirements for existing networks.In the increasingly complex mobile network,the traditional TCP congestion control algorithm can not cope with these complex situations.For example,in some application scenarios,the bandwidth utilization of the network is low or the delay is too large,which directly affects the user's network experience.Therefore,it is especially important to design a TCP congestion control algorithm that adapts to the current cellular mobile network and achieve good results in cellular mobile networks.This paper first introduces the current research status of congestion control algorithms in cellular mobile networks.Then,the existing congestion control algorithms are classified according to their own characteristics,namely delay-based congestion control algorithm,packet loss control algorithm based on packet loss,hybrid drive-based congestion control algorithm and learning-based congestion control algorithm..In this paper,the typical schemes in different categories are selected,and detailed analysis and comparison are carried out.The performance and application scenarios of different congestion control algorithms are analyzed in detail through actual measurement and simulation experiments.Finally,according to the characteristics of the cellular mobile network scenario itself,three key factors affecting the transmission performance are summarized.The performance degradation of various TCP congestion control algorithms in different factors is analyzed and elaborated.Aiming at the problem of transmission performance degradation in cellular mobile networks for various congestion control algorithms,this paper proposes a congestion control algorithm RTT-GRADIENT combined with RTT gradient variation and delay.The algorithm judges the congestion of the current network by delay and RTT gradient change,and adjusts the size of its window according to its gradient change,so as to avoid congestion.After that,the state transition mechanism of RTT-GRADIENT in the Linux kernel and its implementation process are described in detail.Finally,the effects of the experimental platform and the actual test are analyzed and compared.The experimental results show that RTT-GRADIENT can not only achieve better throughput performance under cellular mobile network,but also has the advantage of lower latency.Not only is it more reasonable in terms of latency and throughput balance than the rest of the algorithms.In view of the fact that the DRWA algorithm solves the buffer overflow problem when the throughput in the high-latency network drops drastically,this paper designs a dynamic adjustment receiving window for the Cubic congestion control algorithm to mitigate the problem of buffer expansion and optimize its high-latency network.Throughput performance.Finally,it is more stable and efficient in terms of throughput performance compared to other algorithms.At the end of this paper,the research work is summarized and the working direction of the next research is put forward.