基于多路径TCP拥塞控制的研究

发布时间：2018-05-25 14:23

本文选题：多路径TCP + 拥塞控制　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：近年来,随着互联网传输技术的不断演变,现在的网络通信设备都同时具备多个网络接入端口,这就使得在通信的主机之间可以同时通过多个网络接口,多条路径传输流量数据。这种具有多个网络接口的多宿主特性使得设备具有了更好的可移动性、快恢复能力、安全性和负载共享功能。IETF组织于2009年成立了MPTCP工作组,专门负责多路径传输协议标准的制定,解决传输过程中拥塞控制和路由等问题,发布标准的协议接口,保证与现有TCP协议相互兼容。一条MPTCP连接就是在两个通信主机之间同时创建了多个子流,利用多条子流的并行,实现多路径传输的目的。另外,及时将传输中断链路的数据迁移到其他并行链路可以有效提升网络的鲁棒性,保证数据的可靠性。本文从公平性和均衡拥塞两个方面对多路径拥塞控制策略进行分析。在公平性方面,设计了DWBC(Dynamic Window Based Coupling)算法,该算法能根据公平性的原则,对原有的?因子进行改进,使拥塞窗口能根据链路情况动态变化,通过对共享瓶颈链路的探测来将共享链路来将各子流结合在一起。DWBC能够动态地通过将拥塞窗口不同的子流整合到一个集合中,对同一个集合中的子流使用RTT补偿机制来保证对TCP的公平性。该算法能够适应网络的变化,并能快速从错误中恢复过来。通过在四种仿真场景下进行试验仿真,DWBC都具有不错的性能表现。在均衡拥塞方面,提出了一种基于丢包的DBLIA拥塞控制算法,该算法能够依据现有的拥塞状态进行分类,通过利用拥塞窗口的加速和减速因子,能根据当前网络的拥塞情况来相应的增大和减小拥塞窗口,将数据从高拥塞度子流转移到低拥塞度子流上,能使网络能以最快的速度趋于稳定,并达到最优资源池的状态。并且设定了几个不同的场景和参数进行仿真实验,在相同的场景下与LIA算法进行了比较,发现在加速因子和减速因子的作用下,网络的稳定速度更快,算法更高效。
[Abstract]:In recent years, with the continuous evolution of Internet transmission technology, network communication devices now have multiple network access ports at the same time, which makes it possible to simultaneously pass through multiple network interfaces between communication hosts. Multiple paths transmit traffic data. This multi-host feature with multiple network interfaces enables devices to have better mobility, fast recovery, security and load-sharing capabilities. IETF organized the MPTCP working Group in 2009. It is specially responsible for formulating the standard of multipath transmission protocol, solving the problems of congestion control and routing during transmission, publishing standard protocol interface, and ensuring compatibility with the existing TCP protocol. A MPTCP connection is to create multiple sub-streams simultaneously between two communication hosts. The purpose of multipath transmission is to use the parallelism of multiple sub-streams. In addition, migrating the data of the transmission interrupt link to other parallel links in time can effectively enhance the robustness of the network and ensure the reliability of the data. This paper analyzes the multipath congestion control strategy from fairness and equilibrium congestion. In the aspect of fairness, the DWBC(Dynamic Window Based Coupling) algorithm is designed, which can be applied to the original algorithm according to the principle of fairness. The congestion window can change dynamically according to the link condition. By detecting the shared bottleneck link, the shared link can be combined together. DWBC can dynamically integrate the different sub-streams of the congestion window into a collection. RTT compensation mechanism is used to guarantee fairness to TCP for substreams in the same set. The algorithm can adapt to network changes and recover from errors quickly. DWBC has good performance through four simulation scenarios. In the aspect of equalizing congestion, a DBLIA congestion control algorithm based on packet loss is proposed. The algorithm can be classified according to the existing congestion states, by using the acceleration and deceleration factors of congestion window. According to the congestion situation of the current network, the congestion window can be enlarged and reduced accordingly, and the data can be transferred from the high congestion sub-flow to the low congestion sub-flow, so that the network can stabilize at the fastest speed and reach the state of the optimal resource pool. Several different scenarios and parameters are set for simulation experiments, and compared with the LIA algorithm in the same scenario, it is found that under the action of acceleration factor and deceleration factor, the stability speed of the network is faster and the algorithm is more efficient.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06

【相似文献】