一种基于自适应采样的网络丢包率测量方法

丢包率是基本的网络性能指标之一,现有测量方法在测量精度和探测量开销之间存在冲突.本文提出一种基于自适应采样的丢包率测量方法,利用时延-丢包相关及丢包-丢包相关的特性,根据测量过程中大时延测量样本和丢包事件对网络突发丢包的指示作用,自适应地动态调整探测周期,从而在不大幅提高探测量的前提下,提高了测量准确性.仿真实验结果表明,该方法在测量精度和探测量开销上有较好的折中.与ping相比,该方法测量误差降低较多,而探测量开销则增加不多;与较为准确的BADABING相比,测量误差相差不大,但探测量开销明显减少.

A Network Packet Loss Rate Measurement Method Based on Self-adaptive Sampling

Packet loss rate is one of the basic network performance metrics. However, current measurement methods conflict with measuring precision and costs. A self-adaptive sampling based loss rate measurement method was proposed in this paper. By utilizing the characteristics of delay-loss correlation and loss-loss correlation, it dynamically adjusts probing period adaptively, under the indications of larger delay samples and loss events in measuring process on burst losses. Consequently, it improves the measuring accuracy without the dramatic increase of probing costs. Simulation results have shown that a good trade-off is obtained in this method. When compared with ping, it gets much lower errors with a minor increase of probing costs; and compared with BADABING, the errors are similar, but it greatly decreases probing costs.