【作者】 蒋海；

【导师】 刘淑芬；

【作者基本信息】 吉林大学 ， 计算机系统结构， 2005， 硕士

【摘要】 随着计算机网络技术的飞速发展,电信数据网规模不断扩大、复杂性不断增加以及异构性越来越普遍,基于数据网的各种应用业务也越来越广泛,因此,网络管理的难度也越来越大。近几年,综合数据网络管理系统成为国内电信网管领域的一个关注热点,各运营商在这方面投入了大量的人力和物力。性能监测是综合数据网络管理系统最重要的部分之一,研究性能监测技术的理论和实现机制,具有很大的实际意义;流量预测可以改变传统的响应式网络管理方式,具有较大的研究和应用价值。首先,简单介绍网络管理与SNMP 协议;然后,对性能监测理论及实现方法进行了系统研究和分析,内容包括网络流量监测、利用率监测、可用性监测、端到端性能监测,对不同类型的性能指标,给出了具体实现机制和步骤,并提出一种采集及存储方案,解决了海量性能数据管理的困难;其次,对基于流量模型的网络流量预测技术进行了初步研究,研究分析了常用几种有关自回归的流量模型,并详细给出基于ARMA 模型的流量预测方法,对该模型产生的预测误差进行了分析,给出了我们预测实验结果及其分析,对流量预测技术在网络管理系统中的应用进行了初步探讨,并提出一个预测应用架构;再次,结合“南昌电信综合数据网络管理系统”项目,给出了性能监测的软件设计和系统实现,包括系统功能、体系结构、系统部署、模块设计与实现等主要内容;最后,对我们的研究和实现工作进行简略的总结,并指出我们的下一步研究内容和方向。更多还原

【Abstract】 In this article, the theory and implementation of performance monitoring on telecom data network are systemically studied and analyzed, including network traffic, utilization, availability and end-to-end performance monitoring. To kinds of performance parameters, the scheme and approach of its implementation are presented in detail. A novel method of performance data collecting and storing are also presented. The software design and architecture of one performance monitoring system are described, which is one main parts of the project -Integrated Data Network Management System on NanChang Telecom. On the other hand, the pilot study of model-based network traffic prediction is carried out. The traffic models about autoregressive are analyzed. The ARMA-based network traffic prediction is described and the error of the method is showed, and our experiments in that are presented and analyzed. The application of traffic prediction in network management system is also discussed, and one design framework is given. Performance monitoring is one of the most important parts of network management. By monitoring performance, we get the main values of network parameters, and alerts can be sent to inform system administrator when the values excess aforehand threshold. In terms of performance data, the work state of network can be understood, and its behavior regulations can be found out. It also gives means to plan network, check out and resolve network faults. Traffic monitoring is a fundamental function of network performance monitoring, and the network load can be reflected by monitoring the kinds of traffic information. Resource load can be showed and potential choke point can be found by monitoring the utilization of network devices. It is well known, resources will be wasted in devices with very low utilization, and system performance may decline with very high utilization. By monitoring availability of network element, stability and dependability of network can be denoted. The end-to-end performance, which is the communication performance between a device and one another, including time delay, rate of lost packets and so on, can present the quality of network service. To the all performance parameters mentioned above, the methods of implementations are given in this paper. As is shown in our practical application, the technologies of implementation are of great practical significance, and it provides a good tool for administrator to maintain and administer the network. Our research and implementation are also representative, and the outcomes can be used in telecom data network management in the large majority. In traditional network management, the potential problems will be resolved after receiving the alerts, and it is very possible for the network to have been affected before being settled. This method is called as “response mode”. Such as traffic monitoring, the network often are affected badly when alerts are sent out because of excess of traffic threshold, and it is usually no time to avoid it. The model-based network traffic prediction, in which the traffic model is founded on the old data and the trend of traffic in future is displayed, can meliorate the phenomenon. The possibility of exceeding the threshold will be known through the results of prediction, and the alerts will be sent out if it is large enough. Traffic models are used in predicting the network performance and estimating the control scheme of access. The statistic characteristic of practical traffic can be captured in an accurate traffic model. Because there are strong complexity and isomerism in Internet and behavior of network is high paroxysmal and successional, the Markov-Model and Poisson-Model do not adapt to describe and predict the network traffic. As is shown in a mass of researches, the practical network traffic has self-similarity in statistical sense. Intuitively, a process is self-similar if its statistical behavior is independent of the time-scale. The autoregressive models are analyzed in this paper, including AR, ARMA, ARIMA, F-ARIMA and GARMA. The ARMA-based network traffic prediction is described, and it is shown by the following steps: data collecting, data series smoothing, model establishing, predicting and error analyzing. Our experiments are presented and analyzed based on the practical traffic data. As is shown, the method of prediction can work well. In telecom data network, the network management becomes more and more difficult than before because of the increasing scale and complexity and the expanding application business. The integrated data network management system (IDNMS) is attached importance to apply in the realm of telecom. IDNMS’s management objects involve a large quantity of devices which are notablely different each other. The system functions include network configurations, network performance, network faults, and even some relational application business. The performance monitoring in IDNMS must deal with and store a great lot of data because of a good many monitored devices, and it need provide a friendly and rapid interface to users. In this article, the software design and system implementation of performance monitoring in our project are presented in detail, including system functions, system architecture, system deployment and design of modules. The appraisement of the system is also shown. In application practice, the system’s management objects involving router and switch mainly come from Cisco, Huawei, Ericsson, ZTE, Siemens, and GangWan. As is shown, the system can be used to manage the network efficiently and conveniently. It gives good evidence to the feasibility and validity of our research and design scheme. In the field of network management, much research and system development have been done by many organizations and companies, and更多还原