根据网络上的流统计呈现很强的重尾分布的特性,该文提出了使用周期抽样和counting bloom filter(CBF)技术相结合的方法,即Sample-CBF方法来实现长流识别,并根据抽样策略的不同,将其具体化为两种方法:PSample-CBF方法和FSample-CBF方法。理论分析和仿真结果表明,在存在可容忍流长度测量误差的条件下,两种方法都可以准确识别长流,有效地减少存储空间和提高处理速度。 According to the characteristic of heavy-tailed distribution in network statistics, this paper presents a new method called Sample-CBF, which uses the combination of periodic sampling and counting bloom filter (CBF) The strategy is different and is embodied in two ways: the PSample-CBF method and the FSample-CBF method. The theoretical analysis and simulation results show that both methods can accurately identify the long-stream, reduce the storage space and improve the processing speed effectively under the condition of tolerable measurement error of stream length.