Abstract: In order to improve the performance and reduce the resource usage of flow-based measurement systems，a novel method is presented to collect the data of flow length distributions.First，a packet is sampled from N arrival packets.Then，the counting unit of the packet sampled is updated by two Hash functions.Finally，the statistics stored in counting space are collected periodically to be processed offline.Flow length distributions are obtained by using EM algorithm and least square method.Though testing on different traces，the experiments demonstrate the inferred distributions are accurate.

摘要： 为提高流测量系统的运行效率，减小其所消耗资源，提出了一种新的用于测量流长度分布的估计方法。对到达的报文进行抽样后，用两个哈希函数来确定更新相应计数单元的值，定期收集计数空间中的数据进行离线处理。利用EM算法和最小二乘法，得到了流长度分布。通过应用于来自不同网络的数据进行实验测试，实验结果表明该模型对于流分布的估计是精确的。