The long-range dependence of variable bit rate (VBR) video traffic opens up new areas of research in queueing and performance analysis. Traditional Markov model cannot capture this feature of non-summable autocorrelations and it has been reported that the long-range dependence aggravates the system performance if it is taken into considerations. Futhermore, there is hardly any analytic result of queueing performance when input traffic has both strong short-range correlations and long-range dependence, which is a possible candidate for VBR video traffic model.
We present new analysis of queueing performances by matching of input traffic correlations as exponential function for short-range correlations and hyperbolic decaying function for long-range correlations. With some mathematical proofs, we are able to show that there is a knee point of buffer size below which the buffer distribution is dominated by only short-range correlations and above which by long-range correlations and that the knee point is an decreasing function of utilization parameter. It is shown that if system operates in low utilization and practical buffer size, the long-range dependence of VBR video traffic can be ignored and only Markov model would be sufficient as far as the buffer distribution is concerned.