Analysis Of Characteristics Of End-to-end Path Delay In The Internet

With the rapid development of the Internet, its network size and the number ofusers have been constantly increasing, and its complexity and heterogeneity have beenincreasing, too, which make network management, network security and qualityof service (QoS) facing more severe challenges. Delay is one of the basic parameters fornetwork performance. Understanding the characteristics and laws of network delaysplays a very important role in distributed network monitoring, strengthening networkmanagement and improving service level. However, it is hard to analyze thecharacteristics of network delays because network delays are affected by bursty factorsin the Internet and it remains as an open issue.In this thesis, we focus on the basic status of the changeing of end-to-end delaythrough eliminating the effects of bursty factors. By analyzing the basic status, we canknow its intrinsic characteristics and laws. We propose an analysis method bydecomposing an end-to-end delay into two components: basic delay and delay jitter.Basic delay reflects the overall trend of delay change, and delay jitter reflects theinstantaneous status change in a certain moment. The results of analysis have shownthat most basic delay values follow the normal distribution under normal networkstate. Based on the results, we use basic delay to define the state of path delayin a period. We propose a genetic algorithm to divide the state of path delay and test itsusability. Meanwhile, we make large amount of analysis on the state of path delay usingSpearman’s correlation coefficient for ranked data. The results have shown that thechange of hops and route make little difference to the state of path delay, the delays ofsome individual hops have great influence on the state of path delay. The experimentalresults have shown that this method can reflect the change of network delay well.