互联网IP级拓扑瓶颈时延的研究分析

采用CAIDA提供的海量数据样本，主要针对网络中的瓶颈时延行为特征进行分析.首先针对路径中的关键时延进行统计分析，发现超过80%的路径中产生了瓶颈时延，进一步对瓶颈时延量化分析得出，由于瓶颈时延的存在导致网络直径相差不大的路径，其网络时延相差悬殊；接着将网络时延和瓶颈时延两端的IP地址在地理位置上的分布特征进行统计，发现二者的地理分布特征有较大差别；最后研究分析了产生瓶颈时延的原因，得出在远距离范围内，传播时延是导致瓶颈时延最重要的因素，而在近距离范围内，排队时延是导致瓶颈时延最重要的因素.

Research on the Bottleneck Delay of Internet IP Level Topology

With IP level data collected from CAIDA， the characteristic of bottleneck in Internet was analyzed. Firstly， a statistical analysis about critical delay of network paths was made， which indicated that the bottleneck delay appeared in more than 80% network paths. And then a quantitative analysis on the bottleneck delay showed that the bottleneck made the similar network diameter’s network delay distinct. Furthermore， an analysis on IP addresses of the network delay and bottleneck delay from the degree of the geography location was performed， and it could be concluded that there was a big difference between network delay and bottleneck delay. Finally， a research on the causing of the bottleneck was made， which indicated that propagation delay was the main factor of the bottleneck delay for long distance， whereas the queue delay was the main factor for a short range.