Memory System Architecture Optimization for Enterprise All-RRAM Solid State Drives

Abstract

The Resistive RAM (RRAM) technology is emerging as one of the possible candidates in replacing state-of-the-art NAND Flash for Solid State Drives (SSDs) applications. However, the RRAM architectures developed so far evidence a granularity mismatch between their page size and the typical host application payloads, forcing the use of multi-plane approaches to mimic NAND Flash thus affecting the figures of merit (i.e., bandwidth, latency, and Quality of Service) of a potential "all-RRAM" SSD. In this work we present a RRAM memory system optimization acting both on the internal page size architecture and on the SSD's firmware to find the best configurations able to guarantee the highest performance metrics in enterprise-class SSD applications.