【作者】 刘颖；

【导师】 梁鸿；

【作者基本信息】 中国石油大学 ， 计算机技术， 2009， 硕士

【摘要】 在深入理解自相似,多重分形和尺度行为的基本概念和多种特性以及建模拟合方法后,依据校园网流量数据,对网络中各种形式流量序列进行了分析,利用网络流量尺度行为一般估计法验证校园网络中存在的流量尺度行为,探讨部分利用小波分尺判定法精确分析网络中的普通聚合流量与视频流量,解释当时间尺度趋于微细时出现的多重分形估计方法。在给出网络流量中存在的尺度行为的同时,寻求可能的物理解释,对网络的行为作出合理性的描述,进而分析基于尺度行为的研究新应用点的可行性。在针对校园网络数据处理时,分别从采集方法和流量特性分析作了仔细研究,描述了大时间尺度、小时间尺度和微时间尺度下网络流量所呈现的不同特性。选择了以Libpcap软件分析平台来进行短期流量采集,并在校园网服务器上采用被动监听的方式,选择软件拓展的API编程端口,利用相关分析的编程方法,并设定过滤条件采集短期分析所需数据,并对数据进行进一步的分析处理,统计其表现出来的行为特征。针对校园网采集到的数据进行了网络流量基本特征方面的分析,分别从大时间尺度、小时间尺度和微时间尺度分析了网络流量所体现的特性。发现在不同典型时段的网络流量数据在1ms到150ms(只是大致的数量级)的时间尺度上普遍存在自相似性,而且体现出来的自相似性程度跟网络流量和时间尺度存在一定的关系。网络流量越大时自相似性越明显,网络流量越小时则越小,而在极微小的时间尺度上网络呈现了一定程度的分形。时间尺度取得越大,网络流量所表现出来的自相似性也相应的越大。将网络流量看作一个随机序列,分析网络尺度行为产生的原因。从时间尺度入手,分析并探讨了时间尺度趋于无穷大时的自相似的物理原因,时间尺度趋于微细时的多重分形特性的物理原因。并对现代数据网络中尺度行为原因做了合理性的阐述。更多还原

【Abstract】 Based understanding the basis concept and multi-characteristic of self-similar, multifractal and scaling behavior, and modeling simulate measure, we analysed the all the sequence of network flow of the campus network system. Try to validate existent network flow scaling behavior of the campus network system by using commonly estimate measures. Try to discuss accurate analyse the common aggregation network flow and videl frequency flow by using small wave scaling and determinant measures. Making clear the estimate measure of multifractal, when we study in time-scale going to minuteness. As exist scaling behavior in the network flow, we seek possible physical explaination, and make rationality describing for network behavior. And that, we can analyse the feasibility of new research applications basis scaling behavior.Aiming at dealing data of the campus network flow, we studied how to collect and analyse the characters of network flow, and described the present the different characteristic in the different time-scaling, such as big, small and tiny time-scaling. We choosed Libpcap software analyzing flat to collect the short time flux, and applied the manner of passive monitor in the campus network server. By using the API programming port of the software, and using the interfix analysis programming method, we enacted filter condition to analyse the data we needed. Analyzing and dealing with the data, counting the emerge behavior characteristics.We analysed basic characteristics of network flow in the campus collected data, and analysed in different time-scaling. We find the characteristic of ubquitied self-simular between time-scaling 1ms and 150ms, and which is ralationship with network flux and time-scale. It is more distincting in self-simular, when the network flow becoming more and more big, and it appears multifractal characteristic in the tiny time-scaling by contraries. The time-scaling is more long, the self-simular characteristic is more evidence.By looking upon network flow as a random sequence, analyzing the causation of network flux scaling behavior. Using the time-scaling method, Studied and analyzed the physical reasons of self-simular, when the time-scale is going to infinitiy. And analyed the physical reasons of multifractal characteristic, when the time-scale is going to minuteness. Make out the rationality expatiation in allusion to scaling behavior of modern times data network.更多还原