IP报文总长字段统计属性研究

由于网络技术的迅速发展及吉比特以太网的出现,对流量进行全分组测量越来越困难,采用抽样方法并以抽样样本来估计网络流量的总体特性已成为研究重点.一般情况下,总体服从什么分布是未知的,对总体是否服从某个分布作检验,常用的方法是x^2分布检验法.然而,当样本中IP报文长度比较集中时,大量实验结果表明IP报文的总长字段不服从x^2分布,但并不能认为抽样样本的统计分布和总体的统计分布就不相同.本文提出了IP报文总长字段的一种新的检验方法——累计偏差检验法,并在南华大学校园网总出口上进行了抽样实验,结果证明抽样样本的分布和总体的分布相同,同时也证明了这种检验方法是非常有效的.

Research of Statistical Attribution on IP Message Total Length Field

With the rapid development of network technology and the emergence of the G-bits Ethernet,it is inereasinly difficuh to measure the full-grouping of the Network Traffic.Using the method of sampling to estimate population characteristic of network traffic according to sampling samples becomes the most important research.Generally speakly,it is unknown which distribution population conform to,we often use the test method of x^2 distribution to test whether population conform tocertain distribution.But if the length of IP message appears to be comparatively concentrative and this method is still used to estimate the distribution of the total length of IP message,a grcat number of experiments result manifest that the distribution of the total length of IP inessage does not conform to x^2 distribution,but we can not think that the sampling sample and population do not have the same statistics distributin.Thus,this paper presents a new test method of the total length field of IP message,which is based on a lot of experiments and is called the test method of accumulation deviation.Furthermore,we have done sampling experiment in campus network exit of Nanhua University.The result verifies that the sampling sample and population have the same statistics distribution as well as this test method is very effective.