命中率、字节命中率和延迟时间是Web缓存系统中最重要的性能指标,但是却难以准确、合理地度量不同大小的Web对象的访问延迟。引入字节延迟的概念,为不同的对象延迟建立了一个比较合理的评价标准。提出最小延迟代价的Web缓存替换算法LLC,使用户访问的延迟时间尽可能缩短。实验结果表明,与常用的缓存替换算法相比,LLC算法在有效减少用户感知的访问延迟方面具有较好的性能表现。
Although hit-ratio, byte hit-ratio and access latency are the most important metrics of web cache performance evaluation, they can't beused to measure the access latency of web objects of different sizes quantitatively. The byte latency (BL) is employed and least latency cost web cache replacement policy is presented, which means the lower latency cost, the lower access latency of client. The simulation shows that, LLC outperforms the existed web cache replacement policies in reducing the user perceived access latency.