拥塞控制算法的性能评估及公平性分析

发布时间：2018-09-03 08:35

【摘要】：随着互联网技术的快速发展,当前的网络通信环境也变得日益复杂。在地面互联网中,传输带宽越来越高,链路构成越来越复杂;而在一些特殊网络中,如卫星网络中,通信距离越来越远,传播延时越来越大,这都为当前的TCP传输协议带来了前所未有的挑战。鉴于此,更多的拥塞控制方案被提出使得这个领域成为一个研究热点。本文针对一系列常见的拥塞控制算法(TCP-Reno、TCP-Cubic、TCP-Hybla、TCP-Vegas、TCP-Westwood、TCP-BBR和PCC)进行了理论研究,分析这7种算法针对传统的拥塞控制算法的改进点以及在空间通信中的适应性。在理论分析的基础上,通过搭建地面和空间通信的仿真场景对7种拥塞控制算法进行性能评估。对基于TCP改进的协议我们主要关注的性能参数是发送端的拥塞窗口、慢启动阈值和吞吐量,而对于PCC协议主要关注发送端的发送速率和吞吐量。场景的搭建工作本文通过TC和Netem来动态调整仿真链路的信道参数,信道参数包括往返延时、信道带宽和信道丢包率。往返延时的设定按照地面网络、低轨卫星网络和GEO骨干网络的往返延时来设定。由于要对每个信道参数组合进行多轮仿真测试,为了防止在设定发送端参数和信道参数的过程中出现人为错误因此具体的仿真工作本文主要通过编写自动化仿真测试工具来动态的变更测试场景。仿真工作完成后要根据各算法的性能参数来分析在不同的通信场景中特别是空间通信场景中算法的性能表现为空间通信中的算法选择提供重要参考。同时,新的拥塞控制算法不断被提出,但在全球范围内更多的机器协议栈还构建在传统的拥塞控制算法之上,因此新算法如何和传统的拥塞控制算法保持良好的公平性尤为重要,拥塞控制算法的传输公平性不但影响到现有通信节点能否有效地利用链路带宽,而且注重对算法传输公平性的影响对新算法的普及有重大意义。本文我们通过对新算法和TCP-Reno算法的传输公平性进行测试评估让我们对这些新算法有更清晰的认识。
[Abstract]:With the rapid development of Internet technology, the current network communication environment has become increasingly complex. In the terrestrial Internet, the transmission bandwidth is increasing and the link composition is becoming more and more complex; in some special networks, such as satellite networks, the communication distance is getting farther and farther, and the propagation delay is increasing. This brings unprecedented challenges to the current TCP transport protocol. In view of this, more congestion control schemes have been proposed to make this field a research hotspot. In this paper, a series of common congestion control algorithms (TCP-Reno,TCP-Cubic,TCP-Hybla,TCP-Vegas,TCP-Westwood,TCP-BBR and PCC) are studied in theory, and the improvement points of the traditional congestion control algorithms and their adaptability in space communication are analyzed. On the basis of theoretical analysis, the performance of seven congestion control algorithms is evaluated by setting up simulation scenes of ground and space communication. For the improved protocol based on TCP we mainly focus on the performance parameters of the sender congestion window slow start threshold and throughput while for the PCC protocol we mainly focus on the sending rate and throughput. In this paper, the channel parameters of the simulation link are dynamically adjusted by TC and Netem. The channel parameters include round-trip delay, channel bandwidth and channel packet loss rate. The round-trip delay is set according to the round-trip delay of terrestrial network, Leo satellite network and GEO backbone network. Because of the multiple rounds of simulation testing for each combination of channel parameters, In order to prevent human error in the process of setting the parameters of the transmitter and the channel, the specific simulation work is mainly by writing the automatic simulation test tools to dynamically change the test scenario. After the simulation work is completed, it is necessary to analyze the performance of algorithms in different communication scenarios, especially in space communication scenarios, according to the performance parameters of each algorithm, which provides an important reference for the selection of algorithms in space communication. At the same time, new congestion control algorithms have been proposed, but more and more machine protocol stacks are built on the traditional congestion control algorithms around the world. Therefore, it is very important for the new algorithm to maintain good fairness with the traditional congestion control algorithm. The fairness of congestion control algorithm not only affects whether the existing communication nodes can use the link bandwidth effectively. Moreover, it is of great significance for the popularity of new algorithms to pay attention to the transmission fairness of the algorithm. In this paper, we test and evaluate the transmission fairness of the new algorithm and TCP-Reno algorithm, so that we have a clearer understanding of these new algorithms.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.06

【相似文献】