Measurement and analysis of a large scale commercial mobile internet TV system

Abstract

Large scale, Internet based mobile TV deployment presents both tremendous opportunities and challenges for mobile operators and technology providers. This paper presents a measurement based study on a large scale mobile TV service offering in China. Within the one month measurement period, our dataset captured over 1 million unique mobile devices and more than 49 million video sessions. Analysis showed that mobile viewing patterns are different from that of landline based IPTV and VoD systems. In particular, the average viewing time is significantly shorter, and the channel popularity distribution is more skewed towards top ranked channels than that of landline based systems. For the channel sojourn time, the distribution follows a piecewise model, which combines lognormal and pareto distribution. The lognormal part, which fits the majority of video sessions, more closely resembles the mobile phone call holding time, rather than the power law distribution in the landline IPTV case. In comparing the 3G and WiFi access methods, we found that users exhibit different behaviors when accessing from different networks. In 3G networks, where users are subject to data charge, users tend to have shorter channel sojourn time and prefer lower bit-rate channels. The parameters of the distributions are also different. Understanding these user behaviors and their implications on network traffic are critical for the success of future mobile TV industry.