Internet的访问时间分析  

Analysis on traveling time of Internet

作　　者：李超[1] 赵海[1] 张昕[1] 袁韶谦[1] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《通信学报》2007年第7期54-60,共7页Journal on Communications

基　　金：国家自然科学基金资助项目(69873007)~~

摘　　要：通过CAIDA组织提供的海量样本数据,首先对Internet访问时间进行分析,发现访问时间服从多峰重尾分布。在此基础上对Internet的访问直径与访问时间之间的关系从整体和局部样本进行研究,认为链路延迟是导致相近的访问直径其访问时间相差较大的主要原因,并提出修正算法从样本数据得到链路延迟。对链路延迟样本数据的统计结果表明,超过90%以上的路径其最大的链路延迟占访问时间的1/4以上,并以此定义了支配延迟。最后得出访问时间主要受支配延迟影响,正是由于支配延迟本身导致了相近的访问直径其访问时间存在较大差异的结论。Based on the giant data samples authorized by CAIDA, it found that the Intemet traveling time was typically distributed by multi-modal and heavy-tailed when the Intemet traveling time was statistically analyzed, Furthermore, a research on the relationship between the traveling diameter and the traveling time from the whole and local data samples supposed that link delay was the leading cause for the great variance of the traveling time, and a revising algorithm was proposed to process the raw data samples to obtain link delay. The statistical analysis of link delay presented that the ratio of the biggest link delay to the traveling time in more than 90% of paths was over 1/4.Then a definition of dominant delay was put forward. The traveling time is affected greatly by dominant delay, which makes the similar diameter＇s traveling time distinct.

关 键 词：Internet测量 访问时间 访问直径 链路延迟 支配延迟 

分 类 号：TP393.6[自动化与计算机技术—计算机应用技术]