Approximate Programming Design for Enhancing Energy, Endurance and Performance of Neural Network Training on NVM-based Systems

Ho, Chien-Chung; Wang, Weichen; Hsu, Te-Hao; Jiang, Zhi-Duan; Li, Yung-Chun

doi:10.1109/nvmsa53655.2021.9628582

Cited by 2 publications

(2 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the method proposed by Chen et al [ 14 ], the bit flips were reduced by replacing the newly written value with a floating-point number similar to the old value. Ho et al [ 36 ] proposed an approximate write-once memory (WOM) code to reduce the number of writes to NVM. García et al [ 12 ] proposed a replacement strategy to reduce last-level cache writes to memory and use compression to reduce writes.…”

Section: Related Workmentioning

confidence: 99%

Relieving Compression-Induced Local Wear on Non-Volatile Memory Block via Sliding Writes

Jin

Zhang

et al. 2023

Micromachines

View full text Add to dashboard Cite

Due to its non-volatility and large capacity, NVM devices gradually take place at various levels of memories. However, their limited endurance is still a big concern for large-scale data centres. Compression algorithms have been used to save NVM space and enhance the efficiency of those lifetime extension methods. However, their own influence on the NVM lifetime is not clear. In order to fully investigate the impact of compression on NVM, this paper first studies bit flips involved in several typical compression algorithms. It is found that more bit flips would happen in the shrunken area of a memory block. This induces the phenomenon of intra-block wear unevenness, which sacrifices NVM lifetime. We propose a new metric called local bit flips to describe this phenomenon. In order to relieve the intra-block wear unevenness caused by compression, this paper proposes a sliding write method named SlidW to distribute the compressed data across the whole memory block. We first divide the memory block into several areas, and then consider five cases about the relationship between new data size and left space. Then, we place the new data according to the case. Comprehensive experimental results show that SlidW can efficiently balance wear and enhance NVM lifetime.

show abstract

Section: Related Workmentioning

confidence: 99%

Relieving Compression-Induced Local Wear on Non-Volatile Memory Block via Sliding Writes

Jin

Zhang

et al. 2023

Micromachines

View full text Add to dashboard Cite

show abstract

“…So, researchers come up with different solutions to deal with these limitations, from redesigning the conventional data structures to proposing hardware-level methods, to deploy these new technologies in their systems. Among all the challenges that Energy efficiency [4,9,13,18,[23][24][25][26][27]35] [41, [65][66][67][68][69][70][71] [ [72][73][74][75][76][77][78] NVMs face, in this paper, we focus on (1) low endurance, high energy consumption, and asymmetric read/write related problems and (2) how researchers in different communities, from databases to storage systems to embedded systems and distributed systems, overcome these limitations. Table 3 classifies the research studies from Table 2 based on their memory technologies.…”

Section: Nvm Technologiesmentioning

confidence: 99%

Challenges and future directions for energy, latency, and lifetime improvements in NVMs

Kargar

Nawab

2022

Distrib Parallel Databases

View full text Add to dashboard Cite

Recently, non-volatile memory (NVM) technology has revolutionized the landscape of memory systems. With many advantages, such as non volatility and near zero standby power consumption, these byte-addressable memory technologies are taking the place of DRAMs. Nonetheless, they also present some limitations, such as limited write endurance, which hinders their widespread use in today’s systems. Furthermore, adjusting current data management systems to embrace these new memory technologies and all their potential is proving to be a nontrivial task. Because of this, a substantial amount of research has been done, from both the database community and the storage systems community, that tries to improve various aspects of NVMs to integrate these technologies into the memory hierarchy. In this work, which is the extended version of Kargar and Nawab (Proc. VLDB Endowment 14(12):3194–3197, 2021), we explore state-of-the-art work on deploying NVMs in database and storage systems communities and the ways their limitations are being handled within these communities. In particular, we focus on (1) the challenges that are related to high energy consumption, low write endurance and asymmetric read/write costs and (2) how these challenges can be solved using hardware and software solutions, especially by reducing the number of bit flips in write operations. We believe that this area has not gained enough attention in the data management community and this tutorial will provide information on how to integrate recent advances from the NVM storage community into existing and future data management systems.

show abstract

Approximate Programming Design for Enhancing Energy, Endurance and Performance of Neural Network Training on NVM-based Systems

Cited by 2 publications

References 3 publications

Relieving Compression-Induced Local Wear on Non-Volatile Memory Block via Sliding Writes

Relieving Compression-Induced Local Wear on Non-Volatile Memory Block via Sliding Writes

Challenges and future directions for energy, latency, and lifetime improvements in NVMs

Contact Info

Product

Resources

About