【摘要】 Internet已经成为规模巨大的异构分布式网络,对互联网络拓扑进行判定很困难.现在已有的网络拓扑测量工具主要依赖于所有的中间路由器在测量期间提供的信息.本文分析了大规模异构网络中网络拓扑结构的判定中遇到的问题,研究网络层析成像技术,通过对网络上的主机进行单播的"三明治"测量方案来获得网络节点时延,并给出网络拓扑判定的系统聚类树算法,通过实验验证了系统类聚算法的拓扑数判定结果.在大型网络中使用网络层析成像技术进行网络拓扑推演可行而且具有优势,该方案可以取代TRACEROUTE等工具作为大型网络拓扑判定的解决方案.更多还原

【Abstract】 Today’s Internet is a massive,distributed,heterogeneous network which continues to explode in size.Tools such as TRACEROUTE can not be used because these tools rely on close cooperation from the network internal devices and are incapable of detecting certain types of devices.It is therefore desirable to develop a method for estimating topology that uses only measurements taken at the network edge.Our goal is to identify the logical topology.This article introduces network tomography,a new field which we believe will benefit greatly from the wealth of statistical theory and algorithms.We can not estimate the topology directly,but it is possible to estimate it indirectly.We analyze the related inference technology of Network Tomography,give out the delay-based end-to-end measurement scheme named Sandwich Probe Measurement,which forces us to solely on traffic and queuing characteristics.In our implementation,we use a system clustering tree algorithm to infer network topology,and get a logical topology approximate to real network topology.Our solution will be a feasible solution to supersede other tools like TRACEROUTE.更多还原