Collaborative Research: CNS Core: Large: Systems and Verifiable Metrics for Sustainable Data Centers

合作研究：CNS 核心：大型：可持续数据中心的系统和可验证指标

• 批准号: 2214980
• 负责人: Anshul Gandhi
• 金额: $ 92.84万
• 依托单位: SUNY at Stony Brook
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-10-01 至 2026-09-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2214980&HistoricalAwards=false
• 关键词: Collaborative; Research; CNS; Core; Large

Data centers already contribute significantly to the global carbon footprint. However, the rise in popularity of resource-intensive Big Data and Machine Learning workloads is poised to make data center operations unsustainable. This project designs a suite of Sustainability Aware Software SYstems (SASSY) to enable "sustainable-by-design" data centers. SASSY focuses on sustainability holistically, considering the lifecycle carbon footprint of computing equipment, cleanliness of energy source, and device reliability. To measure per-job end-to-end sustainability costs, a full-stack measurement framework is developed. To involve end-users in sustainability efforts, new programming models and tools are designed to enable users to specify their sustainability and performance objectives. The metrics and models together guide SASSY to make wise data-center-wide sustainable management choices.The adoption of SASSY solutions leads to sustainability savings that benefit the society at large. Further, the SASSY programming models and tools allow developers to build more sustainable applications, enabling "sustainable-by-design" software development. Data center operators and industry partners can directly benefit from SASSY's open-source software and models, which are made public through the project Website: https://www.pace.cs.stonybrook.edu/sassy.html. The next generation of practitioners and researchers are taught to consider sustainability as a first-class metric via educational and mentoring opportunities that the project generates.This project was in response to and partially funded by Design for Sustainability in Computing (NSF-22-060)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.

数据中心已经对全球碳足迹做出了重大贡献。然而，资源密集型大数据和机器学习工作负载的普及将使数据中心运营变得不可持续。该项目设计了一套可持续性意识软件系统（SASSY），以实现“可持续设计”的数据中心。SASSY从整体上关注可持续性，考虑计算设备的生命周期碳足迹，能源清洁度和设备可靠性。为了衡量每项工作的端到端的可持续性成本，开发了一个全栈测量框架。为了让最终用户参与可持续性工作，设计了新的方案编制模式和工具，使用户能够具体说明其可持续性和业绩目标。这些指标和模型共同指导SASSY做出明智的数据中心范围的可持续管理选择。采用SASSY解决方案可以节省可持续性成本，造福整个社会。此外，SASSY编程模型和工具使开发人员能够构建更可持续的应用程序，从而实现“设计可持续”的软件开发。数据中心运营商和行业合作伙伴可以直接从SASSY的开源软件和模型中受益，这些软件和模型通过项目网站https://www.pace.cs.stonybrook.edu/sassy.html公开。通过该项目提供的教育和指导机会，下一代从业者和研究人员被教导将可持续性视为一流的指标。该项目是对计算可持续性设计的回应，并由该项目提供部分资金。（NSF-22-060）该奖项反映了NSF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进行评估，被认为值得支持审查标准。

项目成果

Applied Online Algorithms with Heterogeneous Predictors

• 发表时间: 2023
• 作者: Jessica Maghakian;Russell Lee;M. Hajiesmaili;Jian Li;R. Sitaraman;Zhenhu Liu

SLO-Aware Space-Time GPU Sharing for DL Workloads

DL 工作负载的 SLO 感知时空 GPU 共享

• 发表时间: 2022
• 期刊: Non-archival poster presentation in the 13th ACM Symposium on Cloud Computing
• 作者: Hafeez, Ubaid U.;Gandhi, A.
• 通讯作者: Gandhi, A.

Deep Learning-Assisted Online Task Offloading for Latency Minimization in Heterogeneous Mobile Edge

• DOI: 10.1109/tmc.2023.3285882
• 发表时间: 2024-05
• 期刊: IEEE Transactions on Mobile Computing
• 影响因子: 7.9
• 作者: Yu Liu;Yingling Mao;Z. Liu;Yuanyuan Yang

Predicting Network Buffer Capacity for BBR Fairness

• 发表时间: 2022
• 作者: I. Akgun;Santiago Vargas;Michael Arkhangelskiy;Andrew Burford;Michael McNeill;A. Balasubramanian;Anshul Gandhi;E. Zadok

Online Container Scheduling for Data-intensive Applications in Serverless Edge Computing

• DOI: 10.1109/infocom53939.2023.10229034
• 发表时间: 2023-05
• 期刊: IEEE INFOCOM 2023 - IEEE Conference on Computer Communications
• 作者: Xiaojun Shang;Yingling Mao;Yu Liu;Yaodong Huang;Zhen Liu;Yuanyuan Yang