CAREER: Communication Synchrony

职业：通讯同步

• 批准号: 2047283
• 负责人: Rachit Agarwal
• 金额: $ 78.48万
• 依托单位: Cornell University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-03-01 至 2026-02-28
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2047283&HistoricalAwards=false
• 关键词: CAREER; Communication; Synchrony

Today’s data center networks provide best-effort service—it is impossible to predict the time taken to transmit data across the network. Such unpredictability reduces the set of assumptions that system designers can rely upon from the network, resulting in correctness issues, poor performance and scalability, and high resource overheads. This project aims to design predictable data center networks. Such networks will enable systems that are not only correct by design, but also achieve orders of magnitude better performance and efficiency than today’s systems. Predictable data center networks also have the potential to enable technologies and applications that are infeasible today. This project will also have broader community impact through educational and outreach activities that include creation of publicly available pedagogical materials and tools, and participation in programs that aim to improve the fraction of undergraduates from disadvantaged backgrounds choosing computer science as a major. This project introduces a fundamentally new service model—communication synchrony—that enables data center networks to guarantee bounded, predictable, data transmission times. The project will be organized around three core research thrusts: (1) design and implementation of distributed packet-switched data center networks that enable communication synchrony; (2) development of a theoretical framework to establish fundamental limits, as well as Pareto-optimal designs for networks that enable communication synchrony; and (3) demonstration of the benefits of communication synchrony by designing end-host network stacks that can achieve microsecond-scale latency while scaling to Terabit Ethernet. Communication synchrony has the potential to provide solutions to some of the most difficult and important technical questions surrounding distributed systems, operating systems and end-host hardware for the post-Moore era.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.

今天的数据中心网络提供尽力而为的服务--无法预测通过网络传输数据所需的时间。这种不可预测性减少了系统设计人员可以从网络中依赖的假设集，从而导致正确性问题、性能和可扩展性差以及高资源开销。该项目旨在设计可预测的数据中心网络。这样的网络将使系统不仅在设计上是正确的，而且还将实现比今天的系统高出数量级的性能和效率。可预测的数据中心网络还有可能支持目前不可行的技术和应用。该项目还将通过教育和外展活动产生更广泛的社区影响，其中包括创建公开可用的教学材料和工具，以及参与旨在提高贫困背景本科生选择计算机科学作为专业的比例的计划。该项目引入了一种全新的服务模式--通信同步，使数据中心网络能够保证有限的、可预测的数据传输时间。该项目将围绕三个核心研究主题进行组织：(1)设计和实施可实现通信同步的分布式分组交换数据中心网络；(2)开发理论框架以确定基本限制，以及为实现通信同步的网络进行帕累托最优设计；以及(3)通过设计可实现微秒级延迟的终端主机网络堆栈来演示通信同步的好处，同时扩展到太比特以太网。通信同步有可能为后摩尔时代围绕分布式系统、操作系统和终端主机硬件的一些最困难和最重要的技术问题提供解决方案。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。

项目成果

Understanding host network stack overheads

• DOI: 10.1145/3452296.3472888
• 发表时间: 2021-08
• 期刊: Proceedings of the 2021 ACM SIGCOMM 2021 Conference
• 作者: Qizhe Cai;Shubham Chaudhary;Midhul Vuppalapati;Jaehyun Hwang;R. Agarwal

From Switch Scheduling to Datacenter Scheduling: Matching-Coordinated Greed is Good

• DOI: 10.1145/3519270.3538443
• 发表时间: 2022-07
• 期刊: Proceedings of the 2022 ACM Symposium on Principles of Distributed Computing
• 作者: Rachit Agarwal;S. Rajakrishnan;David B. Shmoys

Towards μs tail latency and terabit ethernet: disaggregating the host network stack

• DOI: 10.1145/3544216.3544230
• 发表时间: 2022-08
• 期刊: Proceedings of the ACM SIGCOMM 2022 Conference
• 作者: Qizhe Cai;Midhul Vuppalapati;Jaehyun Hwang;Christos Kozyrakis;R. Agarwal

dcPIM: Near-Optimal Proactive Datacenter Transport

dcPIM：近乎最佳的主动数据中心传输

• 发表时间: 2022
• 期刊: SIGCOMM
• 作者: Cai, Qizhe;Arashloo, Mina;Agarwal, Rachit
• 通讯作者: Agarwal, Rachit