Popularity-aware cache replacement in streaming environments

With the explosion of Internet streaming applications, system-level support for these applications has become increasingly important. One type of support is caching on streaming servers; this can reduce response latency and server load as well. as improve server throughput. This is because caching obviates the need for disk I/O, which allows more streams to be admitted simultaneously. With the ever increasing sizes of RAM, it is practical to cache part or all of media objects. In this paper we propose a set of file level popularity-based cache replacement algorithms that utilize statically- or dynamically-generated popularity information to direct the replacement of cache blocks. We also consider an environment with both video-on-demand requests and live webcast requests and develop caching strategies for them. We evaluated our popularity-based algorithms extensively under different environments; experiments show that our algorithms improve cache hit rates by around 7% to 14% compared to the most effective current algorithm.