一种基于端到端的网络拓扑推断方法

A Topology Inference Method Based on End-to-end

提出一种新的基于三元分组列车测量拓扑结构的方法,此方法利用叶节点的层析信息将叶节点进行聚类,依据时延抖动和丢包率两个参数计算节点间的相关性,有底向上构造网络拓扑树.该方法较之前方法有效减少了探测包的发送量,并且其推断准确度有所提升,不受到网络负载的影响.最后在NS2仿真环境下进行实验,验证该方法的有效性与准确度,并与其他方法进行比较.

In this paper a new method based on ＂sandwich＂ packet sequences to infer network topology is proposed. The method uses the height of the leaf nodes to cluster, and according two parameters （ delay - jitter and packet loss rate） to calculate the correlation between two nodes, from the bottom- up to construct network topology tree. Compare with the former methods, this method effectively reducing the amount of probe packets, it＇ s inference accuracy improved and not affected by network load. The validity and accuracy of the method are simulated in NS2, and compared with other topology inference methods