An SBI approach to policing MPEG video traffic

This paper presents a framework for dimensioning policing functions based on parametric stationary models of VBR video traffic carried over ATM networks. Parametric models for wide sense stationary processes such as AR and ARMA have been used to accurately model VBR video sources but have been limited to simulation studies because analytical studies of such queueing models are not tractable. This paper investigates the use of a stochastic bounding interval (SBI) approach to characterise VBR video source traffic. Based upon such a characterisation, policing functions with a (σ, ρ) constraint are then relatively easy to define. The approach described is fairly general and can be extended to other parametric models of wide sense stationary processes.