Traffic behavior analysis with Poisson sampling on high-speed network

With the subsequent increasing of backbone bandwidth, the full trace capture will be seriously hindered. The research in the paper results from a study analyzing packet traces obtained from a backbone link in CERNET. Millions of packets are passed per minute through the link that make the traffic measurement difficult or even impossible. The Poisson sampling measure proposed by RFC2330 is used in the research to alleviate the measurement hindering while preserving the statistically representative characteristics of the traffic population. Algorithms for the continuous process sampling and the discrete process sampling are proposed in the paper. The relationship between the mean packet length of the Poisson sampling data sets and the test data set is discussed to show the advantage of the algorithms. The network traffic throughput can be estimated according to the mean length of sampling packets, Poisson rate and during sampling time.