一种基于流数约减的非线性公平采样算法  

Adaptive fair sampling based on reducing flow numbers

作　　者：李海莉[1] 史梦琳[2] 张震[1] 宫阳阳 郭威[1] 王雨[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,郑州450002 [2]郑州电力高等专科学校,郑州450000

出　　处：《计算机应用研究》2015年第6期1826-1829,共4页Application Research of Computers

基　　金：国家"863"计划资助项目(2009AA01A346)

摘　　要：针对现有采样算法存在可扩展性和公平性差的问题,提出一种基于流数约减的非线性公平采样算法(adaptive fair sampling based on reducing flow numbers,AFS-RFN)。AFS-RFN算法首先采用均匀抽样的方法对要统计流数进行约减,获得样本流集合;然后,对属于样本流集合的分组采用非线性的方法进行公平采样,实现控制统计流数目的同时保证统计流信息的准确性。仿真表明,与ANLS(adaptive non-linear sampling)算法相比,AFS-RFN算法大幅降低了存储开销,同时,将算法的公平性提高了60%。算法具有良好的可扩展性和公平性。Since present sampling methods have the shortcomings of non-scalability and low fairness, the paper proposed an algorithm called AFS-RFN. At first, AFS-RFN reduced the flow numbers by using the uniform sampling and got a sample flow set. Then, the packets belonged to the sample flow set were sampled fairly with the non-linear methods. The method controlled the flow numbers to account and guaranteed the accuracy of the information of flows. Compared with ANLS, the simulation demonstrates that the AFS-RFN algorithm saves large amount of memory overhead. At the same time, AFS-RFN improves the fairness by 60% , and has better scalability and fairness.

关 键 词：流量测量 均匀抽样 非线性 公平抽样 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]