Web multimedia files characterization

This thesis introduces multimedia workload characterization embedded in popular Web pages and serving by YouTube. The findings of the study are used to build a multimedia workload generator. The workload generator can be used in simulations that investigates methods in improving caching performance for serving multimedia files, reducing multimedia network traffic, increasing YouTube scalability, decreasing startup delay when playing audio/video files embedded in a Web page or serving by short video sharing services, and evaluating the performance of multimedia servers. In this research, first, an analysis on Web pages consisting of multimedia embedded objects is presented. Also, a characterization study of around 250,000 YouTube popular and regular videos is performed. Based on the analysis of popular Web pages and measurement of YouTube traffic, a workload generator is developed. The workload generator generates the files of popular Web pages and YouTube servers and simulates a user session when accessing a server.