文件名称：流媒体服务器缓存技术研究（硕）
文件大小：1.1MB
文件格式：PDF
更新时间：2011-07-12 05:42:00
VCR VOD 流媒体 缓存
流媒体服务器缓存技术研究，硕士期间的主要研究成果。With the growing popularity of video on demand (VOD) on the Internet, the demands for interactive access to streaming servers have increased dramatically. An impinging problem in VOD systems is how to alleviate the bottleneck in accessing disk I/O or network of limited bandwidth, especially for VCR-like operations. Unfortunately, most of existing streaming proxies are designed for sequential access. In the paper, we propose a novel statistical window caching algorithm for fast interactive access, which uses periodic caching to efficiently manage the cached video data based on the user access pattern. The performance is further enhanced by adopting two policies: prefix caching and variable-sized caching. Simulations using traces from a real VOD server confirm that our proposed method considerably outperforms existing techniques based on uniform segmentation, exponential segmentation, LRU exponential segmentation and adaptive and lazy segmentation algorithm. Under VCR operations, our proposed method achieves a 35% reduction in disk load by using 2.4% of the video’ total size.