An SSMP traffic model based on network measurements

This paper gives an application of the knowledge of the parameters which have the greatest influence on the occupancy and the loss of a finite capacity queue. We show that it is possible to fit the decisive parameters of an SSMP(N) (special semi-Markov process with N states) to those of the measured data, which has been measured on a LAN-network located at the EPFL (Swiss Federal Institute of Technology). The SSMP is a discrete-time process and is a generalization of the Markov modulated Poisson process: each state is characterized by a general probability density function. The example taken to verify this affirmation is the comparison of the expected cell loss in a SSMP(N)/G/1/c queue and the expected cell loss of the converted LAN traffic/G/1/c queue which can be used as model of a LAN-MAN interconnection. The comparisons show that it is possible to correctly model the slotted traffic during a short period of time. >