A study of H.263 traffic modeling in multipoint videoconference sessions over IP networks

This manuscript is a contribution on the modeling of H.263 traffic in multipoint videoconference sessions over IP Networks. Our study includes analysis and assessment of modeling of extensive data gathered during realistic videoconference sessions between commercial H.263-compliant terminal clients (with different videoconference software packages installed). All terminal clients were communicating through a Multipoint Control Unit (software or hardware MCU) at ‘‘switched presence’’ mode and for comparative purposes the same typical videoconference content (a person speaking, with mild movement and occasional zoom/span) was used. Analysis of H.263 data at the frame level suggests that the traffic from the different terminals to the MCU can be represented by a stationary stochastic process with an autocorrelation function (ACF) rapidly decaying to zero and a Gamma formed marginal frame-size distribution (PDF). An accurate analysis of the H.263 traffic from all terminals (with the same visual content and different videoconference software used) shows indicative differences in ACF and PDF of different terminals’ traffic and insights that no generic traffic model can be applied for all cases. Aiming at H.263 traffic modeling, accurate and conservative enough for queuing analysis and network estimation, this study discusses methods for calculating the appropriate model parameters from the observed traffic data and proposes a new technique for unconventional fitting of the PDF. Modeling results indicate the accuracy and conservativeness of the proposed models.