A new model for generating burst traffic based on hierarchical HMM

Bursty has been considered as one of the main features of the Internet traffic. Although many different models have been proposed to profile this feature, most of existing approaches are hardly to simultaneously present the time property and statistic characteristics. In this paper, a new model is introduced for generating burst Web traffic based on hierarchical hidden Markov model. The proposed model includes two underlying Markov state processes. The parent Markov state process is used to describe the large-scale trends or phase of burst traffic. The child Markov process is used to describe the small-scale fluctuations that are happening during a given phase of arrival process. Efficient algorithms for parameter re-estimation and traffic synthesis are derived. Experiments are implemented to validate the proposed model.