基于动态计数型过滤器的网络流公平抽样机制  

FAIR SAMPLING MECHANISM OF NETWORK FLOW BASED ON DYNAMIC COUNT FILTER

作　　者：王宜青[1] 陈庶樵[1] 张震[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《计算机应用与软件》2014年第11期139-142,共4页Computer Applications and Software

基　　金：国家科技重大专项(2009ZX03004-002)

摘　　要：传统的包抽样方法对每一个数据包都以同等的比率抽取,这样就导致了大部分被抽中的是大流,而短流和一般流非常少。高速网络的流量检测需要全面的流信息。针对传统流抽样的缺陷,结合现有的SGS(Sketch Guided Sampling)抽样比与流量成反比的公平抽样思想和动态计数型过滤器,提出更加高效的公平抽样算法DCFS(Dynamic Count Fair Sampling)。DCFS算法使用动态统计过滤器DCF(Dynamic Count Filter)统计流量,相对于SGS算法该方法空间更加高效,而且估计准确性也更好。Traditional packet sampling method draws every packet at the same rate,which results in the vast majority of samples to be drawn are the large flows,but there are few the small and medium flows.Detecting the traffic in high-speed networks needs comprehensive flow information,for the defect of traditional sampling,in combination with the fair sampling idea of existing SGS（ sketch guided sampling）in which the sampling rate is in inverse ratio to traffic and with the dynamic count filter,we propose a more efficient fair sampling algorithm called DCFS（ dynamic count fair sampling）.The DCFS algorithm applies DCF（ dynamic count filter） to count the traffic,and is more efficient in space than the SGS algorithm,its estimation accuracy is also higher.

关 键 词：流统计 布鲁姆过滤器 流量测量 包抽样 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]