节假日对IP骨干网流量的影响研究

发布时间：2018-01-30 20:55

本文关键词： 流量建模时序分析节假日影响　出处：《华中科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：网络流量的时序特征分析是网络测量领域的传统问题。一般认为，大量用户的行为变化可以导致网络流量发生相应的变化。在有些移动通信网络中已经观测到用户每天或者每周行为产生的流量周期性特征。在本文中，我们以一个IP骨干网络的流量作为研究对象，通过时序分析的方法考察中国节假日对流量的影响问题。本文从某电信运营商处获取了一组IP骨干网的为期523天网络流量，包括12个骨干节点的日峰均值。我们开发了一套数据分析算法，试图分析获得骨干节点流量在中国春节和国庆节时的变化特征。我们首先对网络流量数据进行了预处理，通过小波变换和小波变换模极大值法寻找流量数据的奇异点，进而对缺失或者异常的数据进行了修复性的处理。接下来，我们对网络流量进行流量分解，获取其整体增长趋势和周期性波动分量。我们叠加网络的周期性分量获得了各骨干节点的3周和5周的周期流量模板。最后，我们对数据集内的春节和国庆节期间的流量数据和节点模板进行对比，分析节假日期间网络流量的波动情况。结果表明，七天的中国假期会对各节点的常规性周期分量有所影响。进入春节之后，各节点的流量会不同程度的有所下降；究其原因与中国春节的人群迁移性有关，，在节日开始阶段，大部分人可能在忙于与家人团聚，相应的上网流量比平时有所减少。进入国庆节之后，各节点的流量会不同程度的有所上升，但是基本上没有改变流量的每周的周期性流量；究其原因，人们在假日期间有更多的时间可以上网，因此流量呈现较为均匀的提升效果本文的研究结果，一个星期的周期流量的峰值和谷值特征受到国庆节放假的影响不大。
[Abstract]:The temporal characteristic analysis of network traffic is a traditional problem in the field of network measurement. The behavior changes of a large number of users can lead to corresponding changes in network traffic. In some mobile communication networks, the periodic characteristics of traffic generated by users' daily or weekly behavior have been observed. We take the traffic of an IP backbone network as the research object and investigate the impact of Chinese holidays on the traffic through the method of time series analysis. In this paper, we obtain a group of IP backbone network traffic for 523 days from a telecom operator, including the daily peak mean of 12 backbone nodes. We develop a set of data analysis algorithm. This paper tries to analyze the changing characteristics of the traffic of backbone nodes in Chinese Spring Festival and National Day. Firstly, we preprocess the network traffic data. Through wavelet transform and wavelet transform modulus maximum method to find the singular point of traffic data, then the missing or abnormal data are repaired. Next, we decompose the network traffic. We get the whole growth trend and the periodic fluctuation component. The periodic components of the superimposed network are used to obtain the three and five week periodic flow templates of each backbone node. Finally. We compare the traffic data and node templates between Spring Festival and National Day in the data set, and analyze the fluctuation of network traffic during holidays. The results show that seven days of Chinese holiday will have an effect on the regular periodic component of each node. After entering the Spring Festival, the flow rate of each node will decrease to some extent; The reason is related to the migration of people in Chinese Spring Festival. At the beginning of the festival, most people may be busy with family reunion, the corresponding Internet traffic is lower than usual. The flow of each node will increase to a certain extent, but basically does not change the flow of the weekly periodic flow; The reason is that people have more time to access the Internet during the holidays, so the traffic presents a more uniform enhancement effect of the results of this paper. The peak and valley characteristics of a week's cycle flow are not affected by the National Day holiday.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN929.5

【相似文献】