Characterizing user access behaviors in mobile TV system

Video traffic will soon become the majority of the data traffic in mobile networks. However, little is known about how users watch video on mobile devices. Understanding the characteristics of mobile video access can greatly benefit the design of video distribution systems in mobile networks. In this paper, we present an analysis of data logs obtained from a large scale commercial mobile TV system. Our analysis shows that user video accesses can be separated into two distinct categories: browsing and viewing. The user browsing sessions are short video accesses, with its length following a Weibull distribution. The user viewing sessions are considerably longer; its length distribution follows a piecewise model which combines the lognormal distribution and a tail following generalized Pareto distribution. The findings are especially useful for modeling and optimization of Internet based mobile TV systems.