【摘要】 网络速度一直是人们非常关注的问题 ,而包延迟是评价网络速度的一个重要参数 .该文对路由中延迟最大的链路———延迟瓶颈———进行了研究 .文章首先提出了延迟瓶颈的计算方法和两个必要的修正算法 ,通过对世界范围内的 47个节点之间的延迟进行一个月测量和进一步的深入分析 ,得到了如下结论 :(1 )Internet中绝大多数路径的延迟瓶颈消耗了端到端延迟的 1 /3或更多 .(2 )路径中存在多个延迟瓶颈 ,90 %的路径存在一个出现次数超过一半的延迟瓶颈 ,同一路径中不同瓶颈延迟的差异较小 .(3)延迟瓶颈较多出现在传输网 ,且多出现在AS内部 .(4)目前Internet中的绝大多数路径处于轻负载状态 ,瓶颈延迟主要由传播延迟造成更多还原

【Abstract】 This paper presents the concept of delay bottleneck and the algorithm of bottleneck delay, which can locate the bottleneck affecting the end to end delay. Then it gives an analysis of 78357 end to end Internet traceroute measurements conducted between 47 traceroute servers randomly chosen from the Internet. The experiment lasted for 1 month, and 2117 paths were included. It is found that: (1) there exists dominant delay (1/3 of the end to end delay) in most of the routes (90%) in the Internet, and the ratio of bottleneck delay to end to end delay in more than 80% of the paths is between 1/3 and 3/4. (2) There may be more than one delay bottleneck in a network path. 90% of the paths have a delay bottleneck whose frequency accounting for more than half of the total number of traceroutes measuring that path. The difference between bottleneck delays is small. (3) Delay bottlenecks appear in transit network more often, and occur in AS more often. (4) At present, most of the paths in the Internet are lightly loaded, and propagation delay is the major composition of bottleneck delay.更多还原