CAREER: Leveraging physical properties of modern flash memory chips for resilient, secure, and energy-efficient edge storage systems

人和机器创建、捕获、复制和消费的数据呈指数级增长，创造了Zettabytes（10亿TB）时代，对未来的数据存储平台提出了新的挑战。为了满足不断增长的数据存储需求，未来的存储解决方案需要提供更高的位密度、上级性能和能效，以及更好的恢复能力，同时保证最终用户的隐私。解决这些挑战需要在存储系统设计方法中进行范式转变，超越传统的技术不可知的算法方法。为了实现这一目标，该项目将开发利用存储介质丰富物理特性的新技术，从而实现非常规但高效的存储系统。具体而言，该项目（1）开发新的技术特定的表征技术，以探索闪存在运行时的物理特性，（2）通过闪存的逻辑缩放探索新的自适应存储管理技术，以及（3）推进我们对基本存储器阵列特性的了解，如电子噪声，工艺变化，单元退化，以及耐久性和能量精度权衡的限制。此外，该项目还探索了新的数据编码概念，如分数位存储，基于非电荷的数据编码和节能近似存储技术，这些技术对于边缘新兴的数据密集型应用至关重要。这些技术经过量身定制，可确保广泛的适用性、可移植性和易于实施性，有望增强未来各种存储解决方案的弹性、安全性和能效。该项目将在包括边缘/云数据中心和个人存储设备在内的各种存储应用中为消费者、行业和美国政府等提供显著的好处。这些技术将使系统集成商能够在极端环境（例如，核与空间）。此外，将利用工业合作，对本项目中开发的任何相关技术进行关键反馈和可能的商业化。该教育和外展计划包括对研究生和本科生进行网络安全和先进存储技术领域的培训，举办年度研讨会和夏令营以吸引年轻学生，特别是来自代表性不足的群体的学生。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响力审查标准进行评估，被认为值得支持。

Exponential growth in data created, captured, copied, and consumed by people and machines has created the Zettabytes (one billion terabytes) era, posing new challenges to the future data storage platforms. To keep up with growing demands for data storage, future storage solutions will need to provide even higher bit densities, superior performance and energy-efficiency, and improved resilience, while guaranteeing end-user privacy. Addressing these challenges requires a paradigm shift in the storage system design methods beyond the traditional technology-agnostic algorithmic approaches. To achieve this goal, the project will create new techniques that exploit rich physical properties of storage media, thus enabling unconventional but highly effective storage systems. Specifically, the project (1) develops new technology-specific characterization techniques to explore physical properties of flash memory during run time, (2) explores new adaptive storage management techniques through logical scaling of flash memory, and (3) advances our knowledge of fundamental memory-array characteristics such as electronic noise, process variation, cell degradation, and the limits of endurance and energy-accuracy trade-offs. In addition, this project explores novel data-encoding concepts such as fractional bit storage, non-charge based data encoding, and energy-efficient approximate storage techniques that will be critical for the emerging data-intensive applications at the edge. These techniques are tailored to ensure wide applicability, portability and ease of implementation promising to enhance resilience, security, and energy-efficiency of a wide range of future storage solutions.The project will provide significant benefits to consumers, industry, and U.S. government alike in a wide range of storage applications including edge/cloud data centers and personal storage devices. These techniques will enable system integrators to develop near-term disruptive data-intensive applications such as artificial intelligence and predictive analytics at the edge nodes operating in extreme environments (e.g., nuclear and space). In addition, industrial collaborations will be leveraged for critical feedback and possible commercialization of any relevant techniques developed in this project. The educational and outreach plan includes training graduate and undergraduate students in the area of cybersecurity and advanced storage technologies, conducting annual workshop and summer camps to attract young students, particularly from underrepresented groups.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.