SAVI - A Model for Video Workload Generation Based on Scene Length

The development of models for synthetic network traffic generation is essential for performance evaluation of transmission systems. The workload generation models are employed in network simulations, testing and performance predictions. This article presents a new model for workload generation of video encoded with MPEG, called SAVI (Scene Aware Video Workload Generation Model). This model provides an explanation for the short and long range dependence of video traffic, and was designed based on movie scene length. The resulting model presents good possibilities to support generalizations according to the movie style. Besides, SAVI is easier to parametrize if compared with the available models and presents low computational complexity. The model was developed by analyzing the traces of movies publicly available and has been tested through computer simulations. The synthetic traffic generated mimics properly the characteristics of real traffic, including the short and long range dependence.