CAREER: Power and Performance Optimal Distributed Transactions

职业：功率和性能最佳分布式事务

• 批准号: 1824130
• 负责人: Wyatt Lloyd
• 金额: $ 54.19万
• 依托单位: Princeton University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-04 至 2024-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1824130&HistoricalAwards=false
• 关键词: CAREER; Power; Performance; Optimal; Distributed

This project is working to advance the state of the art in the design of storage systems for large-scale websites like Facebook, Uber, and Google that are an integral part of our lives. The project aims to benefit society by making such websites faster, easier to program, and less expensive to build. It will produce research papers that guide industrial efforts and future researchers. It will also produce and release working prototypes of new storage designs that industrial practitioners and other researchers can build off of or compare against. The project will advance discovery and broaden participation by involving undergraduates from the University of Southern California and institutions that serve under-represented groups in research experiences, training graduate students, developing curriculum that integrates teaching and research, and by conducting outreach to K-12 students at schools that serve under-represented groups.The goal of this project is to understand the fundamental tradeoffs between power and performance. The scope of power properties includes consistency models and transactions. The scope of performance properties includes latency, throughput, and their combination. To understand what is not possible we will formulate, conjecture, and attempt to prove impossibility results. To demonstrate what is possible we will design new systems that touch the boundary of our impossibility results. We will implement our designs and publish the implementations publicly. We will evaluate our implementation at large scale to validate they provide the claimed properties and to quantify their benefits and overheads.

该项目致力于推进 Facebook、Uber 和 Google 等大型网站存储系统设计的最先进水平，这些网站是我们生活中不可或缺的一部分。 该项目旨在通过使此类网站更快、更容易编程且构建成本更低来造福社会。 它将产生指导工业工作和未来研究人员的研究论文。 它还将生产和发布新存储设计的工作原型，工业从业者和其他研究人员可以利用这些原型进行构建或比较。 该项目将通过让南加州大学的本科生和为弱势群体服务的机构参与研究经验、培训研究生、开发教学和研究相结合的课程以及向 K-12 学生进行推广来推进发现并扩大参与。该项目的目标是了解权力与绩效之间的基本权衡。权力属性的范围包括一致性模型和事务。 性能属性的范围包括延迟、吞吐量及其组合。为了理解什么是不可能的，我们将制定、推测并尝试证明不可能的结果。为了证明什么是可能的，我们将设计新的系统来触及我们不可能结果的边界。我们将实施我们的设计并公开发布实施情况。我们将大规模评估我们的实施，以验证它们提供所声称的属性并量化其收益和管理费用。

项目成果

Performance-Optimal Read-Only Transactions

性能最佳的只读事务

• DOI: 10.1145/2976749.2978415
• 发表时间: 2024-09-13
• 期刊: Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security
• 作者: Haonan Lu;S. Sen;Wyatt Lloyd
• 通讯作者: Wyatt Lloyd

Regular Sequential Serializability and Regular Sequential Consistency

常规顺序可串行性和常规顺序一致性

• DOI: 10.1145/3477132.3483566
• 发表时间: 2021-09-18
• 期刊: Proceedings of the ACM SIGOPS 28th Symposium on Operating Systems Principles
• 作者: Jeffrey Helt;Matthew Burke;A. Levy;Wyatt Lloyd
• 通讯作者: Wyatt Lloyd

Gryff: Unifying Consensus and Shared Registers

Gryff：统一共识和共享寄存器

• 发表时间: 2024-09-14
• 作者: Matthew Burke;Audrey Cheng;Wyatt Lloyd
• 通讯作者: Wyatt Lloyd