The key to proper resource allocation for mobile video on wireless networks is to have a good model for the resource demands. In this paper, we present the results of analysis of a number of mobile video streams and show that a simple seasonal ARIMA model (SAM) can provide a very good representation for both MPEG4-Part2 and MPEG4-Part10 videos - the formats that are commonly used for mobile videos. The model has been implemented to provide both video frame and RTP packet generators. The model can be used to represent different movies and can be easily adjusted to produce different workloads for simulation studies. We use the SAM generator to compare the performance of three different scheduling methods for video over WiMAX networks: Earliest Deadline First (EDF), Deficit Round Robin (DRR) and a combination of the two. The results show that under overload, EDF introduces unfairness. DRR with deadline is fair and gives the best performance.
Complete paper in Adobe Acrobat format.