Caching Partial Files for Content Delivery

Abstract

Numerous empirical studies have shown that users of video-on-demand platforms do not always watch videos in their entirety. A direct consequence of this is that not all parts of a video are equally popular. Motivated by this, we explore the benefits of dividing files into smaller segments for caching. We treat incoming requests as requests for segments of files and propose a Markovian request model which captures the time-correlation in requests. We characterize the fundamental limit on the performance of caching policies which only cache full files. Next, we propose and analyze the performance of policies which cache partial files. Using this, we characterize the potential for improvement in performance due to caching partial files and analyze its dependence on various system parameters like cache size and the popularity profile of the files being cached.