ALIRS: A High Scalability and High Cache Hit Ratio Replacement Algorithm

Abstract

Nowadays, many systems face to the problem of sharp increase in user data. In order to get efficient access to data, they need a cache system that adopts a replacement algorithm which has both high cache hit ratio and high scalability. However, the mainstream replacement algorithms, which are just like a seesaw between high scalability and high cache hit ratio, cannot get both of the two features satisfactorily. In this paper, we propose a replacement algorithm called ALIRS which is based on LIRS to solve this problem. In our replacement algorithm, the idea is different from the mainstream solutions to improve the scalability. We use parallel computing and Actor concurrency model to hide the expensive cost of replacement operations, which significantly improve the scalability. Furthermore, this idea does not hurt the cache hit ratio of original LIRS at all. We describe the idea and some details in this paper. Experimental results show that ALIRS has both high cache hit ratio and high scalability.