The IEEE Infocom 2013 proceedings have just been made available online, including a contribution by DT / TDG partner on the usage of the basic 2-state Markov models for traffic profiles. The presentation explains how to fully exploit 2-state Markov models for adaptation to autocorrelated processes. As the background, an explicite formula is derived for the 2nd order statisitics of 2-state processes, which covers 2-state semi-Markov processes with four transitions specific traffic rate distributions in the most general format.

 

As the main result, simple 2-state processes are shown to be capable of an amazing good fit of traffic profiles especially with regard to the 2nd order statistics of measured traffic traces, because of a 2-dimensional parameter space is available for adaptation. This means an essential improvement over self-similar and special 2-state models, which are often used but only include a single parameter for fitting the 2nd order statistics, e.g. Markov-modulated Poisson processes (MMPP-2) or Gilbert-Elliot models.