A popularity-driven caching scheme for meta-search engines: An empirical study

Abstract

Caching issues in meta-search engines are considered. We propose a popularity-driven cache algorithm that utilizes both popularities and reference counters of queries to determine cache data to be purged. We show how to calculate query popularity. Empirical evaluations and performance comparisons of popularity-driven caching with the least recently used (LRU) and least frequently used (LFU) schemes have been carried out on collections of real data. In almost all cases, the proposed replacement policy outperforms LRU and LFU.