互联网流采样技术综述

Survey of the Technology of Internet Traffic Sampling

网络流量测量对网络管理、网络监控、网络设计和网络规划具有重要的意义.由于网络带宽以摩尔速率增长,流采样作为一种以有限资源实时获取网络运行状况的方式引起了广泛的关注.文章在分析早期流采样技术现状及不足的基础上,详细综述了大流采样算法、流公平采样算法、全流采样算法及异常检测流采样算法四种互联网流采样技术,对四种技术的性能指标给出了理论分析及证明,最后展望了互联网流采样技术可能发展的四个方向.

Traffic measurement plays an important role in the areas of network management, network monitor, network planning and network scheduling. As the bandwidth growth of lnternet backbone abides by Moore Law, traffic sampling attracts more and more at- tention with the capability of acquiring network operation status at limited cost in real time. Based on the analysis of the current situa- tion and deficiency of early traffic sampling algorithm, four new Internet traffic sampling technologies including elephant flow sam- piing, fair flow sampling, per-flow traffic measurement, and flow sampling oriented on network anomaly detection are surveyed thor- oughly and the performance criterion of the four traffic sampling technologies are analyzed and proved in theory. The four research trends of the technology of Internet traffic sampling are foreseen in the end.