Estimating Properties of Flow Statistics using Bootstrap
Stênio Fernandes, Tatiene Correia, Carlos Kamienski, Djamel Fawzi Hadj Sadok · 2004
Traffic measurement has been gaining increasing attention of the network community in the last years, due to its application in a variety of important areas, such as traffic engineering and network planning. Much effort has been devoted to passive flow measurement since collecting packet-level information in high speed links makes this process extremely complex and expensive. There are some techniques for dealing with flow statistics in current commercial routers and associated measurement infrastructure. However, even though flow-level information is more compact than packet-level information, transmitting and storing it would still impose a significant burden on the operation of a typical Internet Service Provider (ISP). In this paper, we advocate that only a small portion of the flow records need to be preserved for further processing. We propose the use of the Bootstrap resampling technique for deriving statistical properties from a previously preprocessed sampled set of flows. Our results show that only 10% or less of the original sampled statistics is necessary in order for Bootstrap to reconstruct the main characteristics of the original raw flow records.