Optimized Caching Strategy Based on Mutual Dependency of Content and Space Size

Search engines play a crucial role in providing users with the correct answers. Caching techniques are important ways to reduce query latency and increase the hits. In this paper, we propose a cache with three layers to increase hit rates. First, we implement the static filling strategy to fill up the cache. This static strategy considers the mutual dependency between the content of each layer in order to save cache space and store more content in each layer. Second, we propose two dynamic mechanisms to update the cache. Each dynamic mechanism considers the cache size of each entry for the layer where the entry will possibly be added. Then, we evaluate the dynamic mechanism by using different batch sizes. Finally, the cache is assessed by adjusting space size ratios of layers based on the valuable cached contents. We conducted experiments on an AOL query log and an abstract Wikipedia collection to evaluate the proposed cache. The empirical results have shown that our cache outperformed compared methods. Practically, RD-MDC achieved hit rates of 79% and 69% for cache sizes of 5 MB and 3 MB, respectively.