摘要
数据包公平抽样通过牺牲长流的包抽样率以换取更高的短流包抽样率,因而比均匀随机包抽样更能保证数据流之间的公平性.现有的公平抽样算法SGS(sketch guided sampling)存在空间效率低、短流估计误差大的问题.提出了一种空间高效的数据包公平抽样算法SEFS(space-efficient fair sampling).SEFS算法的新颖之处在于采用多解析度抽样统计器对数据流流量作近似估计,各个统计器由d-left哈希表实现.采用在OC-48和OC-192骨干网采集的真实流量数据,在数据流流量测量以及长流检测的应用背景下,对SEFS算法和SGS算法的性能进行了比较.实验结果表明,与SGS算法相比,SEFS算法在空间复杂度降低65%的前提下,仍具有更高的估计精度.特别是对于占网络数据流绝大多数的短流而言,SEFS算法估计精度高的优势更为明显.
Fair packet sampling can obtain a higher packet sampling ratio of short flows by sacrificing the packet sampling of long ones; thus, ensuring better fairness among all flows than uniform random sampling does. However, the previously proposed fair sampling algorithm of Sketch Guided Sampling （SGS） has the drawbacks of poor space efficiency and large estimation error for short flows. In this paper, a space-efficient fair packet sampling （SEFS） algorithm is proposed. The key innovation of SEFS is a multi-resolution d-left hashing schema for flow traffic estimation. The performance of SEFS is compared to that of SGS in contexts of both flow traffic measurements and a long flow identification process that uses real-world traffic traces collected from OC-48 and OC-192 backbone network. The experimental results show that the proposed SEFS is more accurate than SGS in both application contexts, while a reduction of 65 percent in space complexity can be achieved. The improvement of estimation accuracy of SEFS is remarkable, especially for short flows, which comprise as past of a large percentage of whole network traffic flows.
出处
《软件学报》
EI
CSCD
北大核心
2010年第10期2642-2655,共14页
Journal of Software
基金
国家高技术研究发展计划(863)No.2008AA01A323~~
