Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A Primer on Memory PersistencyPersistent Memories

A Primer on Memory Persistency: Persistent Memories CHA PT ER 1 1.1 INTRODUCTION For decades, processor systems have benefited from Moore’s Law, doubling clock frequency andtransistordensityeverytwoyears.Siliconperformanceimprovementhasspurrednumerous innovations in designing faster processing systems, with lower data access latency, at a lower cost,everytechnologygeneration.High-performanceprocessorsystemshavereliedonamulti- tiered memory hierarchy for fast data access. The memory tiers closer to the processors provide faster access latency and higher throughput than the tiers farther from the processor. Harddisksandflashmemoriesoccupythelowesttierinthehierarchy.Theyprovidehigh devicedensityatalowerper-bitcost.Traditionalstoragesystemsarecomposedofharddisksor Flash, as they are non-volatile and can retain data across power failures. However, these mass storage technologies have 100x-1000x higher access latency than SRAMs and DRAMs [1–4]. The static random access memories (SRAM) and dynamic random access memories (DRAM) have been traditionally used to build the processor cache hierarchy and main memories. The SRAM-basedcachehierarchyandDRAMsprovidelowerlatencyandhigherthroughputaccess to application data. Unfortunately, the end of Moore’s Law is near [5]; the technology scaling that the semi- conductorindustryhasenjoyedfordecadesiscomingtoanend.Inrecentyears,DRAMtechnol- ogyhasfacedmajorscalabilitychallenges[6,7]andthecostperGBofDRAMhasplateaued[8]. Meanwhile, the demand for DRAM capacity continues to grow, thanks to the surge in high- performancebig-dataapplications[8].Thetwoconflictingtrendshaveresultedinthedaunting memory scalability challenges in data centers. PersistentmemorytechnologiesaimtoofferasolutionforDRAMscalabilitychallenges. Persistent memories (PMs), also known as storage-class memories or non-volatile memories, are denser and consume less power than DRAM. Several technologies, such as Phase Change Memory, Resistive RAM, and Spin Torque Transfer RAM, have been recently explored for http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

A Primer on Memory PersistencyPersistent Memories

Download PDF
10 pages

Loading next page...
 
/lp/springer-journals/a-primer-on-memory-persistency-persistent-memories-I9I7rwSfhw
Publisher
Springer International Publishing
Copyright
© Springer Nature Switzerland AG 2022
ISBN
978-3-031-79193-2
Pages
1 –11
DOI
10.1007/978-3-031-79205-2_1
Publisher site
See Chapter on Publisher Site

Abstract

CHA PT ER 1 1.1 INTRODUCTION For decades, processor systems have benefited from Moore’s Law, doubling clock frequency andtransistordensityeverytwoyears.Siliconperformanceimprovementhasspurrednumerous innovations in designing faster processing systems, with lower data access latency, at a lower cost,everytechnologygeneration.High-performanceprocessorsystemshavereliedonamulti- tiered memory hierarchy for fast data access. The memory tiers closer to the processors provide faster access latency and higher throughput than the tiers farther from the processor. Harddisksandflashmemoriesoccupythelowesttierinthehierarchy.Theyprovidehigh devicedensityatalowerper-bitcost.Traditionalstoragesystemsarecomposedofharddisksor Flash, as they are non-volatile and can retain data across power failures. However, these mass storage technologies have 100x-1000x higher access latency than SRAMs and DRAMs [1–4]. The static random access memories (SRAM) and dynamic random access memories (DRAM) have been traditionally used to build the processor cache hierarchy and main memories. The SRAM-basedcachehierarchyandDRAMsprovidelowerlatencyandhigherthroughputaccess to application data. Unfortunately, the end of Moore’s Law is near [5]; the technology scaling that the semi- conductorindustryhasenjoyedfordecadesiscomingtoanend.Inrecentyears,DRAMtechnol- ogyhasfacedmajorscalabilitychallenges[6,7]andthecostperGBofDRAMhasplateaued[8]. Meanwhile, the demand for DRAM capacity continues to grow, thanks to the surge in high- performancebig-dataapplications[8].Thetwoconflictingtrendshaveresultedinthedaunting memory scalability challenges in data centers. PersistentmemorytechnologiesaimtoofferasolutionforDRAMscalabilitychallenges. Persistent memories (PMs), also known as storage-class memories or non-volatile memories, are denser and consume less power than DRAM. Several technologies, such as Phase Change Memory, Resistive RAM, and Spin Torque Transfer RAM, have been recently explored for

Published: Jan 1, 2022

There are no references for this article.