Hybrid-LRU Caching for Optimizing Data Storage and Retrieval in Edge Computing-Based Wearable Sensors

Abstract

In the era of the Internet of Things, edge computing-based wearable sensors are rapidly emerging for smart health. The collection, storage, and retrieval of data are the key components of wearable sensors. Therefore, it is important to optimize data storage and retrieval. Phase change memory (PRAM) is a kind of phase change memory that is widely used as a new storage medium. It has the characteristics of nonvolatility, high-density storage. However, it has the disadvantages of asymmetry in reading and writing and limited life. In recent years, PRAM and DRAM were combined into PDRAM as a hybrid memory architecture, to solve the problems caused by PRAM. This paper proposes a new cache policy named hybrid-LRU to adapt PDRAM. Hybrid-LRU uses two different LRU cache policies to distinguish PRAM and DRAM as two different storage mediums. The experimental results show that the hybrid-LRU cache policy improves the performance by 4.2%, and reduces the utilization rate of PRAM in PDRAM by 11.8%. In addition, the energy consumption of writing and reading can be reduced to 87.8%.