Modeling and performance study of MPEG video sources over ATM networks

Slices are the basic resynchronization units of video sources with Motion Picture Expert Group (MPEG) coding scheme and the characteristics of slice provide a better understanding of the correlation property of the video source because the autocorrelation function of the slice-layer data separates the interframe (temporal) and the intraframe (spatial) components of correlation. We propose the correlated state transition (CST) scheme for the modeling of MPEG video source at slice level. For the proposed CST model, each slice forms its state transition independently, however, the state transitions of all slices during traffic simulation are correlated by applying the same random number. The step size of the state transition of each slice is determined by taking the autocorrelation with one slice lag into consideration in order to maintain the histogram of each slice. The results show that the CST model demonstrates a good performance in capturing the characteristics of autocorrelation, which is very important in analyzing network performance, and preserving the behavior of the histogram. Performance of a multiplexer connected with several MPEG video sources are also studied by applying the CST model