SHF:Small: Integrated Hardware-Software Power Regulation, Allocation and Isolation in Consolidated Servers

SHF:Small：整合服务器中的集成硬件-软件电源调节、分配和隔离

• 批准号: 1714389
• 负责人: Anand Sivasubramaniam
• 金额: $ 50万
• 依托单位: Pennsylvania State Univ University Park
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-07-15 至 2023-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1714389&HistoricalAwards=false
• 关键词: SHF; Small; Integrated; Hardware; Software

Multi-megawatt datacenters, while critical to our economy, can pose a severe cost, scalability and environmental burden. Imposed/Voluntary demand response programs, to address these concerns, mandate dynamic power regulation of a datacenter. This requires each of its hundreds of thousands of servers to expose and manage knobs for dynamic regulation. As datacenters resort to increasing levels of application consolidation on high-end servers to reduce their Total Cost of Operation, the co-located applications contend for the finite resources at each server. While prior research has looked at solutions for alleviating this contention for conventional computing resources, co-located applications also contend for the dynamically varying power budget, which is becoming an increasingly scarce resource. This project intends to elevate power to become a first class citizen for explicit management, similar to other hardware resources, in the context of current and next generation servers hosting multiple applications.This project will specifically investigate the following issues: (i) when should server power be regulated? (ii) what amount of the regulated power should be allocated to each application, and recursively to each of its computing resources? (iii) how should power allocations be enforced and isolated? The project will develop evaluation platforms to implement and validate the developed research ideas in each of these "when", "what" and "how" dimensions, using real datacenter/cloud workloads. The research contributions from this project can transform the design of server hardware and software, which make their way into large datacenters. The project proposes curricular enhancements at undergraduate and graduate levels using this research, in addition to the involvement of women, minorities and undergraduate students.

多兆瓦级的电力中心虽然对我们的经济至关重要，但可能会造成严重的成本、可扩展性和环境负担。为了解决这些问题，强制/自愿的需求响应计划要求对数据中心进行动态功率调节。这就要求其数十万台服务器中的每一台都要公开和管理旋钮，以便进行动态调节。随着云计算中心在高端服务器上不断提高应用程序整合水平以降低其总运营成本，位于同一地点的应用程序会争夺每台服务器上的有限资源。虽然先前的研究已经着眼于缓解这种对传统计算资源的争夺的解决方案，协同定位的应用还竞争动态变化的功率预算，该功率预算正成为日益稀缺的资源。本项目旨在提升电源成为一等公民明确管理，类似于其他硬件资源，在当前和下一代的服务器托管多个应用程序的上下文中。本项目将具体调查以下问题：（i）什么时候应该规范服务器电源？(ii)应将多少调节功率分配给每个应用程序，并递归地分配给其每个计算资源？(iii)权力分配应如何执行和隔离？该项目将开发评估平台，以使用真实的数据中心/云工作负载，在“何时”、“做什么”和“如何”维度中实施和验证已开发的研究理念。该项目的研究成果可以改变服务器硬件和软件的设计，使其成为大型数据中心。该项目建议利用这项研究，在大学本科和研究生一级加强课程，并让妇女、少数民族和大学本科生参与。

项目成果

To move or not to move?: page migration for irregular applications in over-subscribed GPU memory systems with DynaMap

• DOI: 10.1145/3456727.3463766
• 发表时间: 2021-06
• 期刊: Proceedings of the 14th ACM International Conference on Systems and Storage
• 作者: Chia-Hao Chang;Adithya Kumar;A. Sivasubramaniam

Getting more performance with polymorphism from emerging memory technologies

• DOI: 10.1145/3319647.3325826
• 发表时间: 2019-05
• 期刊: Proceedings of the 12th ACM International Conference on Systems and Storage
• 作者: Iyswarya Narayanan;Aishwarya Ganesan;Anirudh Badam;Sriram Govindan;Bikash Sharma;A. Sivasubramaniam

Pocolo: Power Optimized Colocation in Power Constrained Environments

Pocolo：功率受限环境中的功率优化托管

• DOI: 10.1109/iiswc50251.2020.00010
• 发表时间: 2020
• 期刊: IEEE International Symposium on Workload Characterization (IISWC
• 作者: Narayanan, Iyswarya;Kumar, Adithya;Sivasubramaniam, Anand
• 通讯作者: Sivasubramaniam, Anand

Mediating Power Struggles on a Shared Server

调解共享服务器上的权力斗争

• 发表时间: 2020
• 期刊: 2020 IEEE International Symposium on Performance Analysis of Systems and Software
• 作者: Narayanan, Iyswarya;Sivasubramaniam, Anand
• 通讯作者: Sivasubramaniam, Anand