CAREER: Enhancing Data Center Storage System with Persistent Memory

持久内存技术被设想为数据中心存储系统的革命。其不可预测的性能、能源和成本优势可能会对数据中心的生产力、效率和利润产生不利影响。尽管前景光明，但这种技术转变可能从根本上改变数十年来内存和存储系统的设计假设，并引入关键的设计挑战。例如，新的性能景观可以改变数据中心的能源效率模型，打破现有的各种系统资源之间的负载平衡；由于不同的失效模式和新存储技术的可靠性问题，传统的数据可靠性方案可能变得次优；当前的应用程序和系统软件可能无法充分利用持久内存提供的好处。本研究的重点是通过充分挖掘持久存储器的潜力来提高数据中心存储系统的性能、能效和可靠性。核心方法是重新思考数据中心架构设计策略和内存/存储堆栈管理，使它们能够感知持久内存。该项目设计了效率建模、关键系统资源负载平衡以及跨单服务器内存/存储堆栈和分布式存储系统的低成本可靠性保证的基本技术。该项目还通过利用持久内存增强了一组关键数据中心应用程序。预计该项目开发的研究理念将显著提高数据中心内存和存储系统的访问性能、能源效率、可靠性和使用寿命，使我们的日常生活更加富有成效。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Persistent memory technique is envisioned to revolutionize data center storage systems. Its unpredictable performance, energy, and cost benefits can adversely affect productivity, efficiency, and profit in data centers. Though promising, this technical transition can fundamentally change the decades-long memory and storage system design assumptions and introduce critical design challenges. For example, new performance landscape can change the energy efficiency model of data centers and break the existing load balance across various system resources; traditional data reliability schemes can become suboptimal due to the discrepant failure patterns and reliability issues of new memory technologies; current applications and system software may not fully exploit the benefits offered by persistent memory. This research focuses on enhancing the performance, energy-efficiency, and reliability of data center storage system by fully exploiting the potential of persistent memory. The central approach is to rethink data center architecture design strategies and memory/storage stack management, making them persistent memory aware. The project devises fundamental techniques in efficiency modeling, load balancing of critical system resources, and low-cost reliability guarantee across single server memory/storage stack and distributed storage systems. The project also enhances a set of key data center applications by leveraging persistent memory. It is expected that research ideas developed in this project will enable significantly improved access performance, energy efficiency, reliability, and lifetime in data center memory and storage systems, taking a large step in making our daily lives more productive.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.