Fuzzy Replacement Algorithm for Cache Memory

Abstract

This paper reports a replacement procedure for cache memories. The procedure is essentially a fuzzy algorithm that makes use of 18 rules to select the cache block to be replaced. These rules are primarily a function of three parameters: age and the frequency of usage of a cache block, and the global hit ratio of the cache system. Computationally, the proposed algorithm calculates a replacement index for each cache block and the block with highest replacement index is selected as a victim. The performance of this fuzzy procedure is compared with traditional replacement algorithms such as least recently used (LRU) and first in first out (FIFO). Our simulation experiments indicate that the proposed algorithm is a strong contender to the traditional counterparts. The unique feature of the proposed algorithm is its flexibility; that is, one can always improve its performance further by fine-tuning the rules.