TES-based video source modeling for performance evaluation of integrated networks

This paper considers modeling methodologies of variable bit-rate (VBR) video sources for performance evaluation of integrated networks. We consider an example in which compressed VBR video is transmitted over a local area network carrying both video and data. We devised a group-of-block (GOB) level simulation of an H.261 algorithm over representative VBR image sequences. This simulation data is processed to obtain a statistical characterization in terms of bit-rate histogram, autocorrelation function, etc. We used a new modeling methodology, called TES (Transform-Expand-Sample), which is a general method for generating autocorrelated time series. TES is a non-parametric method that can accurately capture the histogram and approximate the autocorrelation function of any data set, simultaneously. We used the TES model to drive a network simulation for performance evaluation. The simulation results were compared with those obtained earlier with a simpler frame-oriented autoregressive (AR) model. The numerical results show a small, but not negligible, difference in key network performance measures (such as throughput-delay). NEC USA Inc., C&C Research Laboratories, Princeton, New Jersey 08540 yDavid Sarnoff Research Center, Princeton, New Jersey 08543