摘要
高速链路对实时网络流量监测提出挑战.由于流量采集分析设备性能的限制,采用精确、高效的采样方法进行流量监测分析已成为必然.最简单的固定概率采样能监测较大业务流,但往往忽略掉比例几乎超过80%的较小业务流.数据流算法可以实时高效采集高速链路数据,基于该算法的SGS(sketch guided sampling)采样技术可以实时准确估计流大小分布,但当采样速率增大到监测系统处理能力最大值时,该方法的准确性迅速降低.基于SGS方法,提出一种自适应实时网络流量的采样方法SRGS(sketch and resources guided sampling).该方法将监测系统处理能力作为采样概率调节的一个重要参数.实验结果显示,SRGS方法能够及时根据当前流大小和监测系统处理能力,调节数据包采样概率,准确性高于SGS方法.
The emergency of high speed links brings great challenges on online traffic monitoring and measurement. Due to the capacity restriction of traffic sampling system, an accurate and efficient sampling method is highly demanded. Fixed probability sampling is the simplest technique for detecting bigger traffic flows while discarding the smaller ones which consist almost more than 80 % of the count of whole traffic flows. The smaller traffic flows are vital for the analysis of network traffic. Data streaming algorithm can collect data from high speed links on-line and efficiently. SGS （sketch guided sampling） is based on this algorithm and can evaluate accurately the distribution of flow sizes. But its accuracy declines rapidly when the sampling speed exceeds the capacity of the monitoring system. In this paper, an adaptive sampling method for real time network traffic measurement on high speed links based on the SGS method is proposed, called SRGS （sketch and resources guided sampling）. The SRGS method takes the system capacity as an important parameter to adjust the sampling probability. Experiment results show that the SRGS method can adjust the package sampling probability according to the current flow sizes and the capacity in time. And it is more accurate than the SGS method.
出处
《计算机研究与发展》
EI
CSCD
北大核心
2007年第8期1339-1347,共9页
Journal of Computer Research and Development
基金
国家自然科学基金项目(60403031
90604015)
法国电信研发中心资助项目(46135216)
关键词
流量监测
采样算法
处理能力
业务流
traffic measurement
sampling algorithm
measurement capacity
traffic flow
