CAREER: Coordinated Power Management in Colocation Data Centers

职业：托管数据中心的协调电源管理

基本信息

批准号：
1551661
负责人：
Shaolei Ren
金额：
$ 40.4万
依托单位：
University of California-Riverside
依托单位国家：
美国
项目类别：
Continuing Grant
财政年份：
2015
资助国家：
美国
起止时间：
2015-07-10 至 2022-04-30
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1551661&HistoricalAwards=false
关键词：
CAREER Coordinated Power Management Colocation

项目摘要

CAREER: Coordinated Power Management in Colocation Data CentersThis project focuses on optimizing power management in colocation data centers, which provide facility support to multiple tenants that manage their own servers in shared space. As an integral segment of data center industry supporting the exploding information technology demand, multi-tenant colocation data centers have been massively expanding in both number and scale. The significant concerns are both growing operational costs and environmental issues. Nonetheless, the existing energy-efficiency efforts have been dominantly focused on owner-operated data centers, like Google, whose operators have full and coordinated control over both computing resources and facilities, and hence they are not applicable for colocation data centers in which individual tenants manage their own servers and power consumption in a non-coordinated manner. Consequently, colocation data centers are far behind state-of-the-art owner-operated data centers in energy efficiency and sustainability, which, if left unaddressed, is a major hurdle for sustainable growth in digital economy.To achieve energy sustainability in colocation data centers, this project proposes novel market-based approaches to coordinate individual tenants' power management, in concert with optimizing colocation operator's own server management and non-IT control (e.g., cooling system). This project investigates three complementary research thrusts: (i) Holistically optimizing colocation operator's facility management and economic incentives offered to tenants for energy saving, to limit the usage of carbon-intensive electricity; (ii) Enabling colocation data center demand response to minimize colocation operator's operating cost, increase the adoption of renewable energy, and enhance the power grid reliability; and (iii) Experimentation and validation based on combined software simulation, system prototyping, and real-world implementation.As part of the collaborative efforts from both industry and academia to enable sustainable computing, this project uniquely extends the exploration of power management from owner-operated data centers to multi-tenant colocation data centers. This project can catalyze a shift in future data center management and transform the way that the digital economy evolves, bearing great economic, environmental and societal impacts. This project also contains a significant educational component and provides broad opportunities to attract a diverse population of students to engage in computer science.

职业：托管数据中心的协调电力管理该项目着重于优化托管数据中心的电力管理，这些项目为在共享空间中管理自己的服务器的多个租户提供了设施支持。作为支持爆炸信息技术需求的数据中心行业不可或缺的一部分，多租户托管数据中心在数量和规模上都在广泛扩展。重大关注的是越来越多的运营成本和环境问题。尽管如此，现有的能源效率工作主要集中在所有者经营的数据中心上，例如Google，其运营商对计算资源和设施都具有全面和协调的控制，因此它们不适用于以非协调方式管理自己的服务器和电力消耗的单个租户，在该中心中，他们不适用于托管数据中心。 Consequently, colocation data centers are far behind state-of-the-art owner-operated data centers in energy efficiency and sustainability, which, if left unaddressed, is a major hurdle for sustainable growth in digital economy.To achieve energy sustainability in colocation data centers, this project proposes novel market-based approaches to coordinate individual tenants' power management, in concert with optimizing colocation operator's own server management and non-IT control (e.g., cooling system).该项目调查了三个互补的研究推力：（i）整体优化托管运营商的设施管理和向租户提供节能的经济激励措施，以限制使用碳密集型电力的使用；（ii）启用托管数据中心的需求响应，以最大程度地降低托管运营商的运营成本，增加可再生能源的采用并增强电网可靠性；（iii）基于组合的软件模拟，系统原型和现实世界实施的实验和验证。作为行业和学术界的合作努力的一部分，以实现可持续计算，该项目独特地扩展了从所有者操作的数据中心探索电源管理到多居住验证数据中心。该项目可以催化未来数据中心管理的转变，并改变数字经济的发展方式，并产生巨大的经济，环境和社会影响。该项目还包含重要的教育组成部分，并提供了吸引各种学生参与计算机科学的广泛机会。

项目成果

期刊论文数量（9）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

Increasing the Trustworthiness of Deep Neural Networks via Accuracy Monitoring

DOI：
发表时间：
2020-07
期刊：
ArXiv
影响因子：
0
作者：
Zhihui Shao;Jianyi Yang;Shaolei Ren
通讯作者：
Zhihui Shao;Jianyi Yang;Shaolei Ren

On the Latency Variability of Deep Neural Networks for Mobile Inference

关于移动推理深度神经网络的延迟可变性

DOI：
发表时间：
2020
期刊：
Workshop on Hot Topics in Edge Computing (1-page Poster
影响因子：
0
作者：
Yang, Luting;Lu, Bingqian;Ren, Shaolei
通讯作者：
Ren, Shaolei

Multi-Feedback Bandit Learning with Probabilistic Contexts

DOI：
10.24963/ijcai.2020/427
发表时间：
2020-07
期刊：
影响因子：
0
作者：
Da Yu;Huishuai Zhang;Wei Chen;Tie-Yan Liu;Jian Yin
通讯作者：
Da Yu;Huishuai Zhang;Wei Chen;Tie-Yan Liu;Jian Yin

Your Noise, My Signal: Exploiting Switching Noise for Stealthy Data Exfiltration from Desktop Computers?

你的噪音，我的信号：利用开关噪音从台式计算机中秘密窃取数据？

DOI：
10.1145/3410048.3410094
发表时间：
2020
期刊：
ACM SIGMETRICS Performance Evaluation Review
影响因子：
0
作者：
Shao, Zhihui;Islam, Mohammad A.;Ren, Shaolei
通讯作者：
Ren, Shaolei

Your Noise, My Signal: Exploiting Switching Noise for Stealthy Data Exfiltration from Desktop Computers

你的噪音，我的信号：利用开关噪音从台式计算机中秘密窃取数据

DOI：
10.1145/3393691.3394178
发表时间：
2020
期刊：
SIGMETRICS/Performance Joint International Conference on Measurement and Modeling of Computer Systems
影响因子：
0
作者：
Shao, Zhihui;Islam, Mohammad A.;Ren, Shaolei
通讯作者：
Ren, Shaolei