针对当前分布式抽样方法不能较好地解决抽取样本的随机性和估计无偏性的问题,采取对被抽样的数据包头中没有实际意义的Flag预留位做标记的方法,以保证各测量点抽取样本的一致性;在入口点处采取泊松抽样保证样本的随机性和估计无偏性.理论推导和实验验证均表明:所提方法实现简单、准确性高,抽取样本能较为准确地估计出流量总体特征,并且部署灵活,适合于工程应用.
To address the problem of statistical randomness and unbiased estimate in the distributed sampling,an algorithm was proposed which marked the reserved bit to ensure the consistency of the sample in each measurement point and used Poisson sampling in entry points to ensure the statistical randomness and unbiased estimate.It is proved that the algorithm is simple and high accuracy,and the sample can more accurately estimate the overall traffic characteristics.The algorithm also has flexibility,which is suitable for engineering.