基于分段流行度分布建模的代理缓存空间优化方法

缓存空间大小和命中率是影响代理缓存服务器建设成本和服务质量的重要因素,但两者相互制约,在代理缓存服务器设计时需要折衷考虑。根据均匀分段后的流媒体片段流行度分布,建立代理服务器缓存空间与命中率之间的关系模型,提出了一种基于均匀分段流行度分布建模的代理缓存空间优化方法,能在定价值函数的情况下计算出的最优代理缓存空间大小。实验结果表明,当给定的价值函数中缓存空间大小和命中率的权重相等时,按所提方法优化设计代理缓存空间存储占比比经验设定缓存存储占比为100%、75%、50%、25%和0%时的价值平均提高24%。

Proxy Cache Size Optimization Based on Segmentation Popularity Distribution Modeling

Cache size and hit rate, which affect the price and service quality respectively, are two important factors for proxy cache servers and need to be designed carefully due to their contradiction nature. In this paper, the relation between the proxy cache size and the hit rate for the uniform segmented streaming media is first analyzed. Then a new proxy cache size optimization scheme based on modeling the uniform segmentation popularity distribution is proposed, which can calculate the optimal proxy cache sizes under a given value function efficiently. Experimental results show that for the given value function with equal weights of Cache size and hit rate, the proposed scheme achieves an average 24% improvement of value over the empirical storage proportions of 100%, 75%, 50%, 25% and 0%.