铁路数据网网络资源管理中的网络流量趋势研究

发布时间：2018-05-26 12:14

本文选题：铁路数据网 + NetFlow　；参考：《北京交通大学》2017年硕士论文

【摘要】：随着铁路信息化的快速发展和业务多样化的需要,铁路数据网成为承载铁路各MIS系统数据交互、视频会议、视频监控等通信信息的主要平台,构建在铁路数据网上的应用越来越多,其复杂程度和对网络的依赖程度日益增高。各业务对带宽需求也越来越大,如何更好地了解网络流量状况对网络进行合理的带宽分配成为了铁路数据网网络资源管理研究中的关键问题。基于流属性的网络流量模型是网络性能分析和网络带宽分配的基础,精准的网络流量模型对于业务流量预测、网络拓扑设计、网络性能都有重要的意义。因此,通过统计分析铁路业务流量和各业务子网流量行为特性建立高效的网络流量分析预测模型是实现网络带宽分配的前提,也是实现铁路数据网智能化调配的首要研究课题。但是,目前铁路数据网流量分析停留在简单粗略的监控,流量精确统计、流量预测技术等还有待研究。本文基于网络流量采集以及流量建模的研究,针对铁路数据网承载业务特性,设计基于业务的流量统计方案并且采用分数差分自回归和滑动平均模型对铁路实际网络流量数据进行建模分析和预测。主要研究内容包括以下几个方面:(1)分析归纳了铁路数据网业务承载特性及IP地址分配规律,在此基础上,研究分析相关流量统计技术方法的优劣性,重点分析了 NetFlow技术的数据采集、缓存、老化机制以及聚合策略。设计基于源IP地址前缀匹配聚合策略方法,建立铁路业务系统流量统计方案,为未来基于业务流量预测的带宽分配提供数据基础。(2)根据铁路数据网流量的自相似性和复杂性特点,分析网络流量模型的优缺点,选择分数自回归整合滑动平均(Fractal Autoregressive Integrated Moving Average,FARIMA)模型作为文章的建模分析技术。为了简化FARIMA模型分析算法的复杂度,本文将FARIMA模型分解为差分过程和ARMA过程。通过仿真验证了 FARIMA模型的长相关性,并且对铁路数据网实际流量数据进行FARIMA建模预测。通过与ARMA模型预测拟合对比分析验证FARIMA模型预测的精准度,实验结果证明FARIMA建模预测分析拟合度较高,能够作为铁路数据网网络流量趋势预测分析。本文得出的预测数据能够对网络进行动态带宽分配,对于铁路数据网业务而言,预测出每一个业务流量趋势能够实现各业务子系统VPN带宽分配,在业务网络繁忙时,能够提前预知并且扩大带宽,避免丢包、延时等网络性能的劣化,保障铁路数据网运行安全。在业务网络空闲时,能够合理规划带宽分配,节省带宽资源。
[Abstract]:With the rapid development of railway information and the need of diversification of business, railway data network has become the main platform for carrying the communication information of MIS system, such as data exchange, video conference, video surveillance and so on. There are more and more applications in railway data network, and its complexity and dependence on the network are increasing day by day. The demand for bandwidth of various services is also increasing. How to better understand the network traffic status and how to allocate the bandwidth to the network has become a key problem in the research of network resource management in railway data network. Network traffic model based on flow attributes is the basis of network performance analysis and network bandwidth allocation. Accurate network traffic model is of great significance for traffic prediction, network topology design and network performance. Therefore, the establishment of efficient network traffic analysis and prediction model through statistical analysis of railway traffic and traffic behavior characteristics of each service subnet is the premise of realizing network bandwidth allocation, and is also the first research topic to realize the intelligent allocation of railway data network. However, at present, the traffic analysis of railway data network remains in the simple rough monitoring, accurate flow statistics, flow prediction technology and so on. Based on the research of network traffic collection and traffic modeling, this paper aims at the characteristics of railway data network carrying service. The traffic statistics scheme based on service is designed, and the fractional differential autoregressive model and moving average model are used to model and predict the actual traffic data of railway network. The main research contents include the following aspects: 1) analyzing and summarizing the bearing characteristics of railway data network service and the law of IP address allocation. On the basis of this, the advantages and disadvantages of the related flow statistics techniques are studied and analyzed. The data acquisition, cache, aging mechanism and aggregation strategy of NetFlow technology are analyzed in detail. Based on the source IP address prefix matching aggregation strategy, the traffic statistics scheme of railway service system is established. This paper provides a data basis for bandwidth allocation based on traffic prediction. (2) according to the characteristics of self-similarity and complexity of traffic in railway data network, the advantages and disadvantages of network traffic model are analyzed. The fractional autoregressive integrated moving average Autoregressive Integrated Moving (FRMA) model is selected as the modeling and analysis technique in this paper. In order to simplify the complexity of FARIMA model analysis algorithm, the FARIMA model is decomposed into differential process and ARMA process in this paper. The long correlation of FARIMA model is verified by simulation, and the actual flow data of railway data network is predicted by FARIMA modeling. The accuracy of FARIMA model prediction is verified by comparing with ARMA model prediction and fitting analysis. The experimental results show that FARIMA model prediction analysis has a high fitting degree and can be used as the network flow trend prediction analysis of railway data network. The predicted data in this paper can allocate the dynamic bandwidth of the network. For the railway data network service, it is predicted that each service flow trend can realize the VPN bandwidth allocation of each service subsystem, and when the service network is busy, It can predict and enlarge the bandwidth in advance, avoid the deterioration of network performance such as packet loss and delay, and ensure the safety of railway data network operation. When the service network is idle, the bandwidth allocation can be reasonably planned and the bandwidth resources can be saved.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U29-39;TP393.06

【参考文献】