IPTV increases opportunities of earning profits for ISPs, and opens new content markets to Content Provider (CP) and Program Provider (PP). In order to expand the markets and maximize profits, IPTV providers need to create more appealing programs and better broadcasting schedules by understanding the usage patterns and user behaviours. They can gather log data from the set-top box (STB) and network devices. However, due to huge amounts of log files in large-scale IPTV system, it is not easy to perform deep analysis for various metrics on time. The volume of log data is expected to increase rapidly because of the explosive growth of users. For the efficient management of analysis jobs for diverse metrics under a large set of log data, in this paper, we propose MapReduce-based IPTV log analysis methods on the cloud computing platform, called Hadoop. We carried out experiments with a small testbed and a large one with Amazon service, and found that the MapReduce-based approach outperforms the DBMS by 13 times. From the MapReduce analysis, we also present the characteristics of hourly usage pattern, channel variation, and regional features of IPTV users.
展开▼
