基于被动测量的丢包算法精度分析

东南大学网络空间安全学院 江苏南京 211189

计算机网络; 互联网; 被动测量; 网络性能; 丢包估计; 伪重传

Analysis of accuracy of packet loss estimation algorithm based on passive measurement
LIU Fayuan and DING Wei

College of Cyberspace Security, Southeast University, Nanjing 211189, Jiangsu Province, P. R. China

computer network; internet; passive measurement; network performance; packet loss estimation; spurious retransmission

DOI: 10.3724/SP.J.1249.2020.99025

备注

对边界网中的丢包率进行估计有助于网络管理者更好地获悉评估网络性能.以丢包率为切入点,分析现有丢包估计算法在网络边界应用的可行性,并基于标准数据集进行面向准确度的对比分析.在此基础上,发现现有丢包估计算法产生误差的原因是由于伪重传检测的不准确导致,为此提出针对性的解决方法,即通过伪重传的定义对其进行检测.测试结果表明,改进后的丢包估计算法的准确度得到了明显的提升.将改进后的丢包估计算法运用在CERNET和CERNET2南京主节点的4个校园网中,分别测量这4个校园网的IPv4与IPv6的丢包率.结果表明,IPv6网络中的丢包率低于IPv4网络中的丢包率.

Estimating the packet loss rate in the border network helps network managers better understand network performance and achieve effective estimation of network performance. Taking the packet loss rate as an entry point, the research work analyzes the feasibility of the existing packet loss estimation algorithm applied to the network boundary, and conducts an accuracy-oriented comparative analysis based on a standard data set. On this basis, the author locates the cause of the error in the existing packet loss estimation algorithm and proposes a targeted solution. The test results show that the accuracy of the improved packet loss estimation algorithm has been significantly improved. Finally, the improved packet loss estimation algorithm is applied to the four campus networks of the CERNET and CERNET2 Nanjing master nodes, and the IPv4 and IPv6 packet loss rates of these four campus networks are measured respectively. The results show the packet loss rate in the IPv6 networks is lower than the packet loss rate in IPv4 networks.

·