Buffer occupation probability of trace-driven background streams in hybrid simulation

In this paper, numerical calculation schemes of the buffer occupation probability for a buffer which is fed by a large number of independent and stationary sources are evaluated. The resulting buffer occupation probability density function can be sampled in order to model the impact of a background traffic stream on the foreground traffic in a hybrid fluid-flow packet-event simulation. The calculation methods are based on the Large Deviations asymptotic or on the Central Limit theorem. An estimate combining both approaches is also studied. The so called Moderate Deviations scaling gives a Large Deviations result for a Gaussian source. The paper concludes with a quantitative comparison of the calculation schemes with simulations. Periodic ON/OFF-sources with a heavy tailed distribution are used as arrival processes for performance evaluations. The methods are also applied to the well known Bellcore LAN traffic traces.