<p>Solid-state drives (SSDs) have become the dominant storage solution in modern computer systems, including supercomputers and enterprise data centers, due to their superior speed, energy efficiency, and reliability compared to traditional hard disk drives (HDDs). As data-intensive applications such as big data analytics, artificial intelligence (AI), and high-performance computing continue to grow, the need for larger storage capacities has become critical to handle the ever-expanding volumes of data. To meet this demand, SSDs typically increase capacity in two main ways: (1) by enhancing flash memory density through technologies like MLC, TLC, QLC, and 3D NAND and (2) by expanding the number of channels or flash chips. However, both approaches come with their own challenges. Increasing memory density can lead to reliability concerns, and performance degradation while adding more channels or chips can result in channel contention, limiting scalability. One promising approach to solving this scalability problem is networking flash chips together at the backend of SSDs. However, research into such architectures—viewed as a strong candidate for next-generation SSDs—has been limited by the lack of a unified, accurate, and easily extendable simulation platform. In this work, we introduce MQSimNet, an open-source, integrated simulator for SSDs with networked flash chips, designed for ease of use and minimal effort extensibility.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

MQSimNet: an open-source simulator for next-generation network-based SSDs

  • Amir Mirzaei,
  • Fatemeh Serajeh Hassani,
  • Atiyeh Gheibi-Fetrat,
  • Mina Zabihi,
  • Sina Ghorbani-Jabbedar,
  • Mahmoud Reza Kheyrati-Fard,
  • Ahmad Javadi Nezhad,
  • Seyed Mohammad Hosseini,
  • Negar Akbarzadeh,
  • Jeong-A. Lee,
  • Hamid Sarbazi-Azad

摘要

Solid-state drives (SSDs) have become the dominant storage solution in modern computer systems, including supercomputers and enterprise data centers, due to their superior speed, energy efficiency, and reliability compared to traditional hard disk drives (HDDs). As data-intensive applications such as big data analytics, artificial intelligence (AI), and high-performance computing continue to grow, the need for larger storage capacities has become critical to handle the ever-expanding volumes of data. To meet this demand, SSDs typically increase capacity in two main ways: (1) by enhancing flash memory density through technologies like MLC, TLC, QLC, and 3D NAND and (2) by expanding the number of channels or flash chips. However, both approaches come with their own challenges. Increasing memory density can lead to reliability concerns, and performance degradation while adding more channels or chips can result in channel contention, limiting scalability. One promising approach to solving this scalability problem is networking flash chips together at the backend of SSDs. However, research into such architectures—viewed as a strong candidate for next-generation SSDs—has been limited by the lack of a unified, accurate, and easily extendable simulation platform. In this work, we introduce MQSimNet, an open-source, integrated simulator for SSDs with networked flash chips, designed for ease of use and minimal effort extensibility.