UNIVERSITY OF CALGARY YouTube Workload Characterization by Phillipa Gill A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

Recent developments in Web technology have changed the way people use the Web. This thesis presents one of the first workload characterizations of YouTube, an extremely popular video sharing Web site. Over a three month period, data was collected on campus YouTube usage, including over 23 million transactions. Statistics on the globally popular videos on YouTube were also collected during this time. The characterization presented in this thesis considers properties of YouTube traffic at aggregate, user, and session levels. We find that video content is much larger and has longer transfer durations than traditional Web content. We also observe that the popularity of videos viewed on campus is not strongly concentrated on a small number of videos. At the session level, it is observed that inter-transaction times are impacted by the driving technology of Web 2.0, AJAX. The size of video content also impacts the amount of data sent within user sessions. The results of this thesis are of interest to network and service providers as well as individuals developing models of Web traffic.