A `hi-lo Markov chain model for multimedia traffic in ATM networks

New types of bursty and correlated application traffic, such as variable-bit-rate packet video, require new modeling techniques which go beyond the on-off discrete-time Markov chain model commonly used for voice traffic. The more general hi-lo DTMC model presented increases the applicability of an existing analytic approach developed for a superposition of on-off DTMCs by S.-q Li and J.W. Mark (1990). The hi-lo model allows a wider range of parameters for which one can obtain a closed-form solution for mean queue size, and it provides an upper bound in the case that parameter sets fall outside the solvable range. Specifically, the approach is to scale the previous analysis to larger values of mean traffic rate by having hi and lo states each with a nonzero traffic rate. This is equivalent to an on-off model with a positive offset. This model is used to extend the analytic approach so that there is a solvable range of parameters for any mean, variance and correlation coefficient. In addition, the solvable case is used to develop an upper bound, on mean queue size for cases that do not have a closed-form solution