摘要: We consider a basic content distribution scenario consisting of single origin server connected through shared bottleneck link to number users each equipped with cache finite memory. The issue sequence requests from set popular files, and the goal is operate caches as well such that these are satisfied minimum bits sent over link. Assuming Markov model for renewing we characterize approximately optimal long-term average rate further prove online scheme has same performance offline scheme, in which contents can be updated based on entire files before new request. To support theoretical results, propose an coded caching termed least-recently (LRS) simulate it demand time series derived dataset made available by Netflix Prize. For this series, show proposed LRS algorithm significantly outperforms used (LRU) algorithm.