PFHA: A Novel Page Migration Algorithm for Hybrid Memory Embedded Systems

Abstract

Phase change memory (PCM) and DRAM-based hybrid main memory have emerged as one of the most promising technology for the future embedded systems. Previous research works have focused on page migration strategies to fully exploit the advantages of DRAM and PCM to achieve higher performance and energy efficiency. However, the ability to predict the pages’ future behavior in these strategies could be further improved and most strategies do not consider how to reduce the DRAM energy consumption to further decrease the hybrid system energy. In this article, we propose a novel dynamic hardware migration strategy named periodical frequency hot-region aware page replacement algorithm (PFHA), which utilizes the relation of periodical read and write access frequencies to predict the hotness of a certain page. Meanwhile, PFHA innovatively initiates write hot page self-migration within DRAM to establish a concentrate DRAM write hot region to reduce the DRAM energy. Experiment results confirm that the proposed technique is able to make most of the write hot pages absorbed by DRAM. Meanwhile, it decreases the PCM average access time to a great extent and gets preferable energy consumption.