基于RTT的端到端网络拥塞控制研究
发布时间：2018-10-17 09:31
【摘要】：针对传统TCP协议在无线网络和高带宽延迟网络中的性能下降问题，本文在对RTT分布模型进行研究的基础上，提出了基于RTT的端到端网络拥塞控制方法。 本文首先建立了端到端的网络模型，并在此基础上提出一个基于Gamma分布的统一的RTT分布模型。通过对Internet网络测量获得的RTT数据进行统计分析，发现在网络负载较轻时RTT服从指数分布，在网络负载较重时RTT服从正态分布。仿真结果也表明，无线网络在负载较重时的RTT服从正态分布。理论分析表明，在一定的参数条件下，Gamma分布可以转换为指数分布或正态分布，总之，RTT服从Gamma分布。 其次，本文提出了基于统计过程控制（SPC）方法的端到端网络拥塞控制协议－TCP SPC。该协议将RTT的变化作为网络拥塞状态变化的反馈信息，利用SPC方法处理RTT数据，判断网络负载状态的变化趋势，采取不同的策略调整拥塞窗口的大小。本文在Qualnet仿真软件和Linux操作系统中实现了TCP SPC协议。实验和仿真结果证明，该协议在保证一定公平性的前提下，可以提高无线网络的吞吐率。 最后，本文将PI控制理论用于网络拥塞控制，提出了基于PI控制理论的端到端网络拥塞控制协议－TCP PI。该协议建立了一个端到端的虚拟路由器模型，根据虚拟路由器实时队长与目标队长的差值，利用PI控制理论计算发送端的发送速率，控制路由器队长。该协议利用队长的变化作为网络拥塞的标识，可以避免无线错误丢包对传统TCP性能的影响。仿真结果表明，，该协议能较好的控制路由器队长，提高无线网络和高带宽延迟网络的吞吐率及网络稳定性。
[Abstract]:Aiming at the performance degradation of traditional TCP protocol in wireless networks and high bandwidth delay networks, this paper proposes an end-to-end network congestion control method based on RTT based on the research of RTT distribution model. In this paper, an end-to-end network model is established, and a unified RTT distribution model based on Gamma distribution is proposed. Through the statistical analysis of the RTT data obtained from the Internet network measurement, it is found that the RTT service is exponentially distributed when the network load is light, and the RTT service is normal distribution when the network load is heavy. The simulation results also show that the RTT service of the wireless network is normally distributed when the load is heavy. Theoretical analysis shows that under certain parameter conditions, Gamma distribution can be transformed into exponential distribution or normal distribution, in short, RTT distribution is from Gamma distribution. Secondly, this paper proposes an end-to-end network congestion control protocol, TCP SPC., based on Statistical process Control (SPC) method. The protocol takes the change of RTT as feedback information of network congestion state, processes RTT data by SPC method, judges the changing trend of network load state, and adopts different strategies to adjust the size of congestion window. In this paper, the TCP SPC protocol is implemented in Qualnet simulation software and Linux operating system. Experimental and simulation results show that the proposed protocol can improve the throughput of wireless networks on the premise of ensuring certain fairness. Finally, PI control theory is applied to network congestion control, and an end-to-end network congestion control protocol, TCP PI., is proposed based on PI control theory. The protocol establishes an end-to-end model of virtual router. According to the difference between real-time and target length of virtual router, the transmission rate of the sender is calculated by using PI control theory, and the length of the router is controlled. The protocol uses the change of queue length as the symbol of network congestion, which can avoid the influence of wireless error packet loss on the performance of traditional TCP. The simulation results show that the protocol can control the router length and improve the throughput and network stability of wireless networks and high bandwidth delay networks.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN915.04
本文编号：2276237
[Abstract]:Aiming at the performance degradation of traditional TCP protocol in wireless networks and high bandwidth delay networks, this paper proposes an end-to-end network congestion control method based on RTT based on the research of RTT distribution model. In this paper, an end-to-end network model is established, and a unified RTT distribution model based on Gamma distribution is proposed. Through the statistical analysis of the RTT data obtained from the Internet network measurement, it is found that the RTT service is exponentially distributed when the network load is light, and the RTT service is normal distribution when the network load is heavy. The simulation results also show that the RTT service of the wireless network is normally distributed when the load is heavy. Theoretical analysis shows that under certain parameter conditions, Gamma distribution can be transformed into exponential distribution or normal distribution, in short, RTT distribution is from Gamma distribution. Secondly, this paper proposes an end-to-end network congestion control protocol, TCP SPC., based on Statistical process Control (SPC) method. The protocol takes the change of RTT as feedback information of network congestion state, processes RTT data by SPC method, judges the changing trend of network load state, and adopts different strategies to adjust the size of congestion window. In this paper, the TCP SPC protocol is implemented in Qualnet simulation software and Linux operating system. Experimental and simulation results show that the proposed protocol can improve the throughput of wireless networks on the premise of ensuring certain fairness. Finally, PI control theory is applied to network congestion control, and an end-to-end network congestion control protocol, TCP PI., is proposed based on PI control theory. The protocol establishes an end-to-end model of virtual router. According to the difference between real-time and target length of virtual router, the transmission rate of the sender is calculated by using PI control theory, and the length of the router is controlled. The protocol uses the change of queue length as the symbol of network congestion, which can avoid the influence of wireless error packet loss on the performance of traditional TCP. The simulation results show that the protocol can control the router length and improve the throughput and network stability of wireless networks and high bandwidth delay networks.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN915.04
【参考文献】
相关期刊论文 前2条
1 刘颖;张宏莉;李东;张冬艳;;测量的Intnernet链路延迟建模[J];哈尔滨工业大学学报;2009年10期
2 毕经平,吴起,李忠诚;Internet中的包延迟分布与包丢失关系研究[J];计算机工程;2002年10期
本文编号：2276237
本文链接：https://www.wllwen.com/kejilunwen/wltx/2276237.html
