随机分组抽样下流大小分布估计

流大小分布是网络测量中一个重要的度量。已有的研究表明在MLE（极大似然估计）算法中运用TCP流的协议信息能够更好的估计流大小分布。本文详细比较了运用TCP流的SYN包和TCP序列号信息的几种MLE算法，并在此基础上结合实际应用提出了一种对小流采取细粒度、对大流采取粗粒度的非均匀粒度的流大小估计算法。实验结果表明，该算法在减少了MLE估计计算量的同时，提高了粗粒度后大流估计精度。

Flow Size Distribution Estimation from Random Packet Sampling

The flow size distribution is an important metric for network measurement. Previous work has shown that using TCP protocol information improves Maximum Likelihood Estimator (MLE) to estimate the flow size distribution. We compare several MLE algorithms that using SYN flag information and TCP sequence numbers. And then we present a non-uniform grianed estimator, which relies on that many applications require fine grained estimates of smaller flow sizes, and require coarse grained estimates of larger flow sizes. Experiments are conducted with two real network traces. Results show that our approach leads to a speedup in the computation, and provides more accuracy for coarse grained estimates of larger flow sizes.