找回密码
 立即注册
流量traffic抽样The | OA系统 2022-06-26 78 0star收藏 版权: . 保留作者信息 . 禁止商业使用 . 禁止修改作品
流量工程和网络行为学研究涉及到网络运行状态的测量、建模、特征化和控制,通过这些方面的研究能够掌握并优化网络运行状况,增强网络运行的可靠性,更为有效的进行网络的规划和设计。网络流量监测是流量工程的一个重要内容,通过流量监测获得各种流量的统计信息是网络管理,网络服务分析,安全和错误监测,网络计费等应用和研究的基础。随着近年来网络流量爆Z性的增大,以抽样的方式进行流量监测已成为流量监测技术的一个重点内容。本文在对流量监测技术相关理论详细了解的基础上,对现有的流量直接捕获,RMON,NETFLOW,SFLOW等流量监测技术,在流量信息获取,信息传送等方面进行了详细的分析和比较。论文还分析了流量监测的多种抽样方法,将其归纳为传统抽样,轨迹抽样,自适应类抽样三类,认为传统抽样更适用于流量监测,然后结合实际流量数据,对几种常用传统抽样方法在数据包长分布,协议分布,流量吞吐量,IP流量统计等流量统计信息方面的表现进行了讨论。实验数据分析表明,在相同条件下不同抽样方法的表现区别不大,而同一种抽样方法在不同流量统计信息上的表现不同。对于数据包长分布和协议分布,由于其精确度要求不高,抽样数据结果能够很好的满足要求,由抽样数据计算得到的流量吞吐量也相当精确,但误差随抽样比率的提高增长较快,而对于IP流量统计,抽样数据结果的误差比前三种相对较大。

(The research of traffic engineering and network behavior involves the measurement, modeling, characterization and control of network operation status. Through these studies, we can master and optimize the network operation status, enhance the reliability of network operation, and plan and design the network more effectively. Network traffic monitoring is an important part of traffic engineering. Obtaining the statistical information of various traffic through traffic monitoring is the basis of network management, network service analysis, security and error monitoring, network billing and other applications and research. With the explosive increase of network traffic in recent years, traffic monitoring by sampling has become a key content of traffic monitoring technology. Based on the detailed understanding of the relevant theories of traffic monitoring technology, this paper makes a detailed analysis and comparison of the existing traffic monitoring technologies, such as traffic direct capture, RMON, NetFlow, sFlow, in terms of traffic information acquisition and information transmission. The paper also analyzes various sampling methods of traffic monitoring, and summarizes them into three categories: traditional sampling, trajectory sampling and adaptive sampling. It is considered that traditional sampling is more suitable for traffic monitoring. Then, combined with the actual traffic data, the performance of several commonly used traditional sampling methods in traffic statistics such as packet length distribution, protocol distribution, traffic throughput and IP traffic statistics is discussed. The analysis of experimental data shows that different sampling methods have little difference in performance under the same conditions, while the same sampling method has different performance on different traffic statistical information. For packet length distribution and protocol distribution, due to their low accuracy requirements, the sampled data results can well meet the requirements, and the traffic throughput calculated from the sampled data is also quite accurate, but the error increases rapidly with the increase of the sampling ratio. For IP traffic statistics, the error of the sampled data results is relatively larger than the first three.)

[下载]13525784617.rar




上一篇:电脑端口监控
下一篇:科来网络分析系统技术交流版2010.