Least Recently Used Caching in the Presence of Statistical Locality

Least Recently Used cache replacement rule is at the core of many replacement algorithms in modern distributed networks. The wide popularity of this rule is primarily due to its high performance, self-organizing nature and ease of implementation. Therefore, in order to further the insights of designing network caching protocols, it is important to gain a thorough understanding of the baseline LRU replacement policy. In particular, it is essential to characterize its performance in the presence of statistical locality in document request sequences [1].