The Research And Analysis On The Delay Characteristics In Internet Macroscopic Topology

Along with scientific-technological progress and high-speed economic development, the Internet has become an indispensible part of human society. And the network performances of the Internet gradually draw the attention of more and more researchers. Delay as an important influencing factor of the network performances has also gradually become a research hotspot. Elaborately analysing the behavioural characteristic of delay is conductive to find network bottleneck and optimize the network configuration. And combining with statistical methods, it is possible to further discover the distribution characteristics and influencing factors of the delay.This work is mainly to analyse the behavioural characteristics of delay, and the abundant data of IPv4-level used in this paper is collected by CAIDA. Analysis on the characteristics of network deley was given first. It is found that the network delay is typically distributed by multi-modal and heavy tailed, and this phenomenon is not accidental. And three main factors which may affect the network delay were studied, the results show that there is no relationship between network delay and traveling diameter; And analysis from the angle of end to end distance supposes that the propagation delay generated by the long-distance is an important factor; Research on the max link delay presents that 90% paths have bottleneck, and max link delay is the most important factors of network delay for the paths with a network delay greater than 85ms. Therefore that bottleneck delay is defined to study the network delay.The characteristics of bottleneck delay were also studied, the results indicate that bottleneck delay leads to the great difference of network delay while the network diameter is approximate. Furthermore, it is found that bottleneck tends to appear in the middle parts of paths. Compared with the distribution features of the network delay, bottleneck delay mainly distributed within one country, especially dstribute in the same place; And bottleneck delay tends to appear in the same AS autonomous domain. By the way, link delay weighted average algorithm was wrote to extract the new bottleneck delay, the result indicates that the new and original bottleneck delay appear at the same location in more than 80% of the paths, and the value of new and original bottleneck delay change relatively little; Besides, the geographical distribution characteristics of new and original bottleneck delay show little change.Analysis on the evolutionary character of bottleneck delay in the last five years shows that both the average of network delay and bottleneck delay increase year by year; The proportion of bottleneck delay in the network delay keeps increasing with years; And the relative position of bottleneck delay is more and more close to the middle part of the paths; Besides, bottlenecks are more likely to occur within one country, with less happening in the intercontinental link; The evolution characteristics of bottleneck delay in the AS autonomous domain shows that nearly two-thirds of bottlenecks distribute in the same AS autonomous domain every year.