Emerging NVM

Abstract

There has been a surge of interest in Non-Volatile Memory (NVM) in recent years. With many advantages, such as density and power consumption, NVM is carving out a place in the memory hierarchy and may eventually change our view of computer architecture. Many NVMs have emerged, such as Magnetoresistive random access memory (MRAM), Phase Change random access memory (PCM), Resistive random access memory (ReRAM), and Ferroelectric random access memory (FeRAM), each with its own peculiar properties and specific challenges. The scientific community has carried out a substantial amount of work on integrating those technologies in the memory hierarchy. As many companies are announcing the imminent mass production of NVMs, we think that it is time to have a step back and discuss the body of literature related to NVM integration. This article surveys state-of-the-art work on integrating NVM into the memory hierarchy. Specially, we introduce the four types of NVM, namely, MRAM, PCM, ReRAM, and FeRAM, and investigate different ways of integrating them into the memory hierarchy from the horizontal or vertical perspectives. Here, horizontal integration means that the new memory is placed at the same level as an existing one, while vertical integration means that the new memory is interleaved between two existing levels. In addition, we describe challenges and opportunities with each NVM technique.

Survey of Storage Systems for High-Performance Computing

“…As stated in [4], the upcoming non-volatile memory will probably revolutionize the memory stack. The technologies will be used to add new memory layers (vertical integration) and to support existing technologies (horizontal integration).…”

“…Large-scale observation systems (for example, in astrophysics) as well as small-scale data loggers (for example, Internet of Things) will require storage systems that can consume and store data at unprecedented scale, ideally with in-transit data processing capabilities. In many cases, scientific workloads are lacking market relevance, and are thus not a priority for many 4 Only individual records are self-describing. 5 Depends on the chosen data format.…”

“…With 3D NAND, the capacity of SSDs will improve further, which might eventually replace HDDs in data centers [41]. NVRAM will likely have a considerable impact on the storage landscape, but most of the known candidates are not ready for mass production or remain too expensive to replace other technologies [4]. Many data transformations such as compression or encryption could be performed out-of-core, either on GPUs or in-transit [2,42].…”

Energy-Efficient Memory Mappings based on Partial WCET Analysis and Multi-Retention Time STT-RAM

“…Store (0) in block 22 creates a lifetime that spans the entire duration of the program because it is defined at the beginning and read in each iteration of the loop. Conversely, the lifetimes defined by stores (1) and (2) in block 22 only reach the first iteration of block 23 in which they are redefined. The same holds for the data lifetimes defined at stores (3) and (4) in block 23, alive only across one iteration.…”

“…In the current study, we consider a specific emerging NVM, named Spin-Torque Transfer random access memory (STT-RAM) [2]. While such a technology has a quasi-null leakage, one major challenge concerns the mitigation of its high write latency and energy cost.…”