Datacenters have enriched the planet, making services like e-commerce and web search available to billions. However, the growing carbon footprint of datacenters adds to climate change and, in the long term, could harm society's perception of the whole field. Our ongoing work builds mechanisms for sustainable datacenters, i.e., datacenters that profit while maintaining low carbon footprints. Intellectually, the work contributes new analytic models that relate a datacenter's carbon footprint to its utilization and energy costs. These models must consider complicated issues like solar- and wind-energy outages, heavy-tail and diurnal e-commerce workloads, and workload and energy balancing within datacenters (e.g., dynamically powering servers off and DVFS). Further, the models must be holistic, explaining these factors in terms of profit. In terms of broad impacts, this work will improve the economic competitiveness and sustainability of datacenter-driven computing.

Project Report

We devised power delivery techniques to power servers with renewable energy. We enabled energy accounting via power delivery devices, e.g., grid ties, power delivery units, and transfer switches. Intellectually, our power delivery systems increased the percentage of renewable energy routed to specific servers by an order of magnitude. A picture of the renewable-powered cluster is attached. We published and demoed this system. [deng-noms-2012(x2), deng-issst-2011 (award paper)]. We also explored the use of renewable-energy accounting in Internet services. An undergraduate thesis [Bunch, 2012] provided the initial framework for an email system that exposes carbon footprint to end users. In practice, green hosting services expose carbon footprint to end users by offering zero-carbon or negative-carbon footprint hosting packages. We found green hosting increases a hosting sites revenues, despite making the site more costly to operate. Using data from real hosting sites and our renewable powered cluster, we proposed an approach called Adaptive Green Hosting. In Adaptive Green Hosting, green hosts adjust their carbon offsetting policies to maximize their profit. This paper was published in ICAC, a premiere forum for adaptive computing systems [deng-icac-2012]. As of this writing, it is one of the most cited works from ICAC 2012. In terms of broader impact, the papers published have been cited often and provided motivation for larger computing systems. We also initiated a seminar series with Morehouse College. In 2010-2012, PI Stewart presented this ongoing work and exposed students to graduate school. In 2012, a student from Morehouse joined Ohio State University with a research fellowship. The grant funded PhD, masters and undergraduate students. Most notably, Nan Deng published 2 first-author conference papers and 3 poster/workshop papers with funding from this grant. References: MantisMail: Green Content Delivery for Email, Nan Deng, Chen Li, Christopher Stewart, and Xiaorui Wang Poster at USENIX Conference on Networked Systems Design and Implementation, Lombard, IL, 2013 Caching for Sustainability, Alexander Bunch and Christopher Stewart, Undergraduate Thesis, The Ohio State University Technical Report, Columbus, OH, June 2012 Adaptive Green Hosting, Nan Deng, Christopher Stewart, Jaimie Kelley, Daniel Gmach and Martin Arlitt, International Conference on Autonomic Computing, San Jose, CA, 2012 Policy and Mechanism for Carbon-Aware Cloud Applications, Nan Deng, Christopher Stewart, Daniel Gmach, and Martin Arlitt, Network Operations and Management Symposium, Maui, Hawaii, 2012 Carbon-Aware Cloud Application, Nan Deng Network Operations and Management Symposium, Maui, Hawaii, 2012 Concentrating Renewable Energy in Grid-Tied Datacenters , Nan Deng, Christopher Stewart, and Jing Li, International Symposium on Sustainable Systems Technology, Chicago, IL, 2011

Agency
National Science Foundation (NSF)
Institute
Division of Computer and Network Systems (CNS)
Type
Standard Grant (Standard)
Application #
1230776
Program Officer
M. Mimi McClure
Project Start
Project End
Budget Start
2012-07-01
Budget End
2014-06-30
Support Year
Fiscal Year
2012
Total Cost
$200,000
Indirect Cost
Name
Ohio State University
Department
Type
DUNS #
City
Columbus
State
OH
Country
United States
Zip Code
43210