Design space exploration of latency and bandwidth in RRAM-based solid state drives

The continuous request for higher storage density in Solid State Drives (SSD) is pushing the NAND-Flash technology to their reliability and performance limits. Among many memories technology candidates to replace them the Resistive RAM (RRAM) concept seems to emerge. However, before designing an entire SSD based on RRAM memory devices it must be performed a design space exploration of the disk features as a function of the memory and disk architecture. In this work, we perform an exploration of RRAM-based SSD as a function of the RRAM page size and of the disk ability to serve multiple outstanding read commands. The bandwidth figures and the statistical assessment of the SSD latency show that in corner conditions these figures become comparable to that of a NAND Flash-based SSD, even if with an accurate host system co-design the RRAM solution can clearly outperform the NAND Flash technology. The continuous request for higher storage density in Solid State Drives (SSD) is pushing the NAND-Flash technology to their reliability and performance limits. Among many memories technology candidates to replace them the Resistive RAM (RRAM) concept seems to emerge. However, before designing an entire SSD based on RRAM memory devices it must be performed a design space exploration of the disk features as a function of the memory and disk architecture. In this work, we perform an exploration of RRAM-based SSD as a function of the RRAM page size and of the disk ability to serve multiple outstanding read commands. The bandwidth figures and the statistical assessment of the SSD latency show that in corner conditions these figures become comparable to that of a NAND Flash-based SSD, even if with an accurate host system co-design the RRAM solution can clearly outperform the NAND Flash technology