基于背景流量的故障检测研究.docVIP

 基于背景流量的故障检测研究# 林博1，徐明伟2，李琦2** （1. 北京邮电大学网络与交换国家重点实验室，北京，100876； 5 10 15 20 25 30 2. 清华大学计算机科学与技术系，北京 100084） 摘要：随着端到端服务的日益增多，端到端服务的可靠性也越来越重要，链路正常与否是影 响其可靠性的重要因素之一。有研究表明链路故障每天都在发生，现有的基于 Traceroute 的故障检测一般都是采用周期性地向目的地址发送探测报文，导致探测报文的开销过大，故 本文提出了基于背景流量的 Traceroute 的故障检测算法(Traceroute-based Failure Detection Algorithm with Background Traffic，TFD 算法)。当网络存在实际流量时，以 背景流量代替探测报文实施对网络故障的检测，并引入 GEANT 以及 Abilene 的真实网络流量 对该算法的性能进行实验和分析，结果表明在有背景流量的情况下，该算法的探测报文开销 有了明显的降低。 关键词：故障检测；可靠性；流量感知；Traceroute 中图分类号：TP393 A Study of Background Traffic Aware Failure Detection LIN Bo1, XU Mingwei2, LI Qi2 (1. State Key Lab of Networking and Switching Technology, Beijiing University of Posts and Telecommunications, Beijing 100876; 2. Department of Computer Science，Tsinghua University，Beijing 100084) Abstract: With the continuous increasing of end-to-end service, the reliability of the end-to-end service is becoming more and more important. Link failure is one of the most serious problems which impact the reliability. Investigations show that Link failures happen every day, while existing failure detection based on Traceroute are generally used to send probe packets to the destination periodically. With these general methods, the overhead of probe packets would be huge. This paper presents Traceroute-based failure detection algorithm with background traffic (TFD).When there is background traffic in the network, we would choose using background traffic instead of probe packets to detect the failure. We also introduce the traffic matrix of GEANT and Abilene to check and analysis the performance of the algorithm. The results show that the overhead of probe packet in our method has been significantly reduced when there is background traffic in the network. Key words: Failure Detection；Reliability；Traffic Aware；Traceroute 35 0 引言 链路是网络中很重要的组成部分。链路故障可能会导致网络节点便成为一个个孤岛。如 果链路发生故障，网络中通信的带宽可能随之降低，网络的通信能力也会下降；另外链路故 障之后，路由收敛的时间一般在 2s-10s 不等[1]，在路由重新收敛之前的这段时间内，用户的 40 请求或数据无法按照原来的路由正确到达故障链路的下一跳，此时这些已发出