SPX: Integrating Persistent Memory in the Cloud

云中的海量数据和高计算强度的大规模应用需要大数据中心的数千台机器。此外，对于新的数据密集型应用对更快、更节能且可扩展的性能的需求也在不断增加。不幸的是，随着技术规模的放缓，半导体行业一直面临着在处理大数据集的同时提供更好的性能和降低功耗的重大挑战。为了提供更好的性能和更低的成本来处理具有严格延迟约束的海量数据的云应用，服务提供商正在转向内存中框架来存储工作数据。通过探索新兴内存技术的作用，该研究项目具有提高云计算性能的潜力。这项研究中提出的想法将弥合架构、系统和软件工程社区之间的差距，并将支持系统支持和自动化工具，以适应持久云中的应用程序。该项目最终将实现一个完整的持久云系统，使得云应用可以透明地适配，而不需要大量程序员的努力。本工作的目标是以跨系统堆栈的整体方式启用持久云系统，使得持久云应用可以在没有大量程序员的情况下在系统中适配。努力。为了设计持久云系统，这项工作是通过三个主要研究方向提供从应用程序到硬件的全栈支持，这三个主要研究方向需要解决：(I)对持久存储系统的轻量级存储层支持，(Ii)基于应用程序特征和NVM中的权衡的数据监控和放置，以及(Iii)应用程序级的自动化持久支持。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The massive volume of data and high computing intensity of large-scale applications in the cloud require thousands of machines in big data centers. In addition, there is an increasing demand for faster, energy-efficient, and scalable performance from new data-intensive applications. Unfortunately, as the technology scaling slows down, the semiconductor industry has been facing a major challenge in providing better performance and reducing the power consumption while processing large datasets. To provide better performance and lower costs for cloud applications that manipulate massive data with tight latency constraints, service providers are moving towards in-memory frameworks to store the working data. By exploring the roles of emerging memory technologies, this research project has the potential to improve cloud computing performance. The ideas developed in this research will bridge the gap between architecture, systems, and software engineering community and will enable system support and automated tools for adapting applications in the persistent cloud. The project will eventually enable a holistic "persistent cloud system" such that the cloud applications can be adapted transparently without significant programmers' effort.The goal of this work is to enable a persistent cloud system in a holistic manner across the system stack such that the persistent cloud applications can be adapted in the systems without significant programmers? effort. In order to design a persistent cloud system, this work is to provide full stack support from the applications to hardware through three major research directions that need to be addressed to design a full-stack persistent cloud system, (i) lightweight storage layer support for persistent memory systems, (ii) data monitoring and placement based on application characteristics and trade-offs in NVM, and (iii) automated persistency support at the application-level.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.