中文摘要：

针对流媒体自身内在流行度变化的特征，提出了一种新的基于两点流行度的缓存算法，根据片段所在的不同区域，分别采用不同的片段流行度计算标准对片段流行度进行估算，进而选择流行度高的片段进行缓存．利用实际用户访问数据，将该算法与等长分段的缓存算法和指数分段的缓存算法进行了比较，仿真结果证明该算法可以在获得与这些算法相近的请求延迟率的条件下，取得最高的字节命中率．

英文摘要：

The characters of video internal popularity from .a real trace were analyzed and a novel caching scheme is proposed based on the popularity at the beginning point and steady point of the video. For the different position of the segment, there are different schemes to decide its popularity. The segment with high popularity will be chosen to cache. In the simulation, a trace driven method was used according to the real trace data. Simulation results demonstrate that, compare to the uniform segmentation caching algorithm and the exponential segmentation caching algorithm, the proposed caching scheme can achieve high byte hit rate and similar request delay rate in the same conditions.