一种面向Web内容分发的缓存替换策略

A novel cache replacement policy for Web content delivery

研究了Web内容分发和对Web服务性能有重要影响的缓存替换机制。考虑到当前的缓存替换策略主要采用基于频度与本地局部性的替换基准，而且通过对实际Web数据访问情况的分析发现访问间隔的变化率对命中率的影响具有更高的准确性，提出了一种基于访问密度与大小混合的缓存替换策略。该策略通过统计近期缓存对象的平均访问间隔，结合该对象的字节大小进行缓存替换，并分别在固定对象数目空间与对象字节空间上进行对比实验，实验结果表明，该策略比最近最少使用（LRU）和最近频繁使用（LFU）算法提升3％-5％的命中率，比空间与频度混合贪心（GDSF）算法提升5％～8％的字节命中率。

The Web content delivery and the cache replacement policy greatly affecting the performance of Web server were studied. Considering that the current cache replacement schemes mainly use the frequency and locality as the basis of replacement and it is found the access interval change rate is more valuable in predicting the new objects arrival through analyzing the real network logs, a novel cache replacement policy based on the access density and the object size was proposed. When using this novel method, the cache can achieve a higher byte hit ratio. The experimental results shows that the method improves 3 %- 5 % of the hit rate than the LRU （ least recently used） and LFU （least frequently used）, and 5%- 8% of the byte hit rate than the GDSF （reedv dual size and freouencv.