Modelling, parameter assessment and multiplexing analysis of bursty sources with hyper-exponentially distributed bursts

This paper has a two-fold objective: first, to develop a method for modelling and parameter assessment of ON-OFF traffic streams exhibiting a hyper-exponential ON-period distribution and, second, to analyse the multiplexing behaviour of such streams, with emphasis on cases where two different time scales (i.e. very short and very long bursts) are present. Traffic of such characteristics arises in many practical applications, like concentration and shaping of independent packet streams, Variable Bit Rate Video with frame refreshment, etc. The proposed model is a three-state Markov Modulated Rate Process (MMRP), the parameters of which are directly assessable by measurements on the real traffic stream, while analytical formulae are provided for the case of shaped data streams. A Near-Complete-Decomposability (NCD) method, developed recently for fluid-queueing models, is reviewed and, then applied to the proposed model for streams possessing divergent time scales. Specific examples demonstrate the applicability of the proposed method on a wide range of traffic streams.