Modelling of variable bit rate signals using burstiness in ATM networks

This paper focuses on the use of the burstiness parameter to model video signals in ATM networks. There has been a considerable amount of work on traffic models using on-off processes with independent distributions which are deemed to be approximate to the traffic in ATM networks. Much of this work has concentrated on the cell level instead of the burst level to formulate the model. Asynchronous transfer mode has been researched in many different areas to cover different topics such as bandwidth allocation, call admission control etc. As the original work was with telephone traffic, most of the traffic models were inherited from telephone systems using the Poisson process in a continuous format instead of a discrete format. These types of models were regarded as simple but were not accurate enough to reflect the traffic in high speed networks such as the ATM. Extensive research has been done using the Markov modulated Poisson process or the Markov modulated Bernouilli process. Their analytical models can characterize the burstiness and correlation assuming the arrival processes are either exponentially or geometrically distributed. This type of model is considered in order to determine the traffic characteristics in terms of the probability distribution function, mean, moment, burstiness and correlation. The paper presents the results and draws a conclusion on this model which is more appropriate to deal with burst type traffic