An approximation for the loss probability in a multiserver queue fed by a shifting level process

A shifting level (SL) process is useful to traffic modeling because it is very easy to match the stationary distribution and autocorrelation function of the SL process with those of measured real traffics. Especially, SL processes have been used to traffic modeling for VBR video traffics reflecting the long range dependence and the scene structure. We give an approximation for the loss probability in a multiserver queue fed by a SL process. Numerical examples show that our approximation is very accurate.