Analysis of Flow Inter-Arrival Time Distributions

E. Daskalova, M. Ilvesmäki, and R. Kantola (Finland)


IAT distributions, flow classification, traffic measurements, Quality of Service, and Internet


Flow identification is one of the main blocks for modeling Quality of Service (QoS) in the Internet for the purpose of supporting various real time applications that have high quality requirements. Therefore, it is essential to understand flow level statistics and their behavior. This paper considers flow inter-arrival time (IAT) distributions as a whole and application by application. Analysis starts by distinguishing applications carried by the traffic mix and finding the most popular ones. Those applications are further studied by plotting histograms for presenting IAT distributions superimposed with exponential distributions and by applying goodness-of-fit statistical modeling. The IAT distribution analyses indicate that it may be feasible to divide traffic flows into two classes by observing the length of tails of the histogram distributions. Further research is still needed with more recent traffic traces and applying different measurement based mechanisms.

Important Links:

Go Back