Non-uniform Segmentation Strategy Of Streaming Media Based On IPTV User Behaviors

Proxy cache technology can solve streaming media service’s problems of startup latency, consume too much bandwidth,etc,satisfy it’s high quality requirements of low jitter, low latency,etc,by cache the most popular streaming media content on the edge of network and provide the nearest service for the user. But streaming media content often has the characteristics of the great amount of data and take up too much cache space, segmentation strategy that can enhance the cache performance become one of the important research content of streaming media proxy cache technology.This paper collect five months, up to 3 million real user access log of China telecom’s large IPTV system which based on CDN, including 260000 video, users of this system has reached 2 million, and the various aspects of user behavior is analyzed, and finally puts forward two kinds of non-uniform streaming media segmentation strategy.The main work is as follows:1. According to the actual users access data of the IPTV system based on CDN, on-demand users concurrent access time distribution, play time, the external and internal popularity distribution of on-demand video and video segments is analyzed, and the characteristics of user access is researched as well.2. In the understanding of the user behavior characteristics, this paper researches the relationship between segmentation and the cache hit rate, and relationship between the external popularity of on demand videoand the number of cache replacement. according to the relationships, this paper present a non-uniform segmentation strategy of streaming media based on popularity classifying, this segmentation strategy use a finer granularity where cache replacement is frequent, and use segmentation granularity coarser in other videos, will make full use of the cache space.The experimental result of real users access data shows that the proposed segmentation strategy can improve byte hit rate and reduce the total number of segments which can save segmentation management cost of system.3. This paper has a further study of the influence of popularity on cache hit rate, and analyzes the relationship between the storage rate and the lower limit of popularity of the cache boundary, and presents a segmentation strategy of streaming media based on the best segmentation point predicting, this strategy ensure that each video is divided into at most two segments according to the estimated segmentation point, the experimental result shows that the proposed segmentation strategy can improve byte hit rate and reduce the total number of segments,which are beneficial to improving proxy server cache performance.