基于Netflow的网络流量测量基础设施建设  

An infrastructure for flow-level network-wide traffic measurement based on Netflow

作　　者：高磊[1] 杨家海[1] 张辉[1] 李福亮[1] 张宾[1] 

机构地区：[1]清华大学信息网络工程研究中心,北京100084

出　　处：《广西大学学报（自然科学版）》2011年第A01期78-82,共5页Journal of Guangxi University（Natural Science Edition）

摘　　要：随着互联网主干网带宽的不断升级,直接对IP分组数据进行采集、存储和分析会产生巨大的测量开销,因而针对流数据的流量测量越来越受到关注。在流级别的测量工具和协议中,Netflow由于其出色的兼容性和易于部署的特点而得到广泛应用。传统基于Netflow的流数据测量系统往往是集中式的,在全网范围内缺乏必要的协作机制,因而极易受到负载分布不均衡、可扩展性差等问题。引入P2P的设计思想来实现均衡负载,并提供高度可扩展性,另外基于对实际流量数据的观察,提出了在IPv6环境下P2P流量识别的方法。The fine-grained flow level measurement is getting increasing demand in recent years. Though it fails to be a generic solution for its biased sampling, Netflow is promising for its compatibility with prevalent touters and its convenience to perform direct flow level measurement of both IPv4 and IPv6 traffic. Traditional flow level measurement systems based on Netflow are mostly centralized and each of them independently performs traffic analysis of its local flow records without any coordination in a large-scale network, suffering from unbalancing workload and bad scalabili- ty. In this paper we present the construction of Flowlnfra which is a scalable infrastructure for network-wide flow measurement of pure IPv6 flow records from Netflow v9 exports. Through the assessment of its performance and flexible features, we show that FlowInfra achieved enhanced ability and robustness to perform network-wide flow level measurement. In addition, based on the observation on CERNET2＇s traffic data, we propose a new method to identify P2P traffic over IPv6.

关 键 词：P2P 分布式计算 NETFLOW 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]