HQ: An Architecture for Web Cache Replacement Algorithms in Distributed Systems

Abstract

Distributed cache is capable of accelerating the process of retrieving an enormous amount of data. In order to optimize the cache performance in distributed environment, we present an architecture of cache replacement algorithms. The architecture called HQ (Hybrid Queue) focuses on organizing distributed cache clusters hierarchically, making use of the web data distribution. In addition, we also introduce a new cache replacement algorithm PDE (Predictive Dead-time Elimination), based on a compound measurement of data cached with the second exponential smoothing technique. Experiment indicates that HQ could bring 3% increment on cache hit-rate on average, while PDE could make a 3% increment as well, comparing with the most widely used algorithms, such as LRU, FIFO, LFU, LRFU, and EXP1 etc.