
Recent studies estimate that data centers — which house thousands of computing servers and equipment — are responsible for about 2 percent of total greenhouse emissions, equivalent to the emissions of the entire airline industry. Even worse, those emission levels are expected to increase to 5 to 7 percent, and government and funding agencies are urgently calling for solutions.
A team of faculty at Stony Brook University, along with colleagues at Binghamton University and Pennsylvania State University, are tackling this challenge with the help of a $1.5 million National Science Foundation (NSF) Large Computer and Network Systems (CNS) grant to study sustainability in data centers.
Anshul Gandhi from the Department of Computer Science leads a team of faculty including Dongyoon Lee (Computer Science), Zhenhua Liu (Applied Mathematics and Statistics), Shuai Mu (Computer Science) and Erez Zadok (Computer Science), on the project titled “Systems and Verifiable Metrics for Sustainable Data Centers.”
SBU researchers are focusing their efforts on sustainability costs. Data centers, such as those operated by Google, Amazon and Microsoft, have miles of space for IT equipment that allow them to provide services such as Google Docs, Dropbox and MS Office 365. However, providing these services to customers worldwide comes at the environmental cost of a staggering release of carbon emissions.
“One of the key hurdles to sustainable computing is the lack of metrics to measure sustainability costs. If we want to optimize something, we have to be able to accurately measure it first,” said Gandhi.
“In addition to being accurate, it is also vital that sustainability costs can be verified and audited by multiple independent entities, to prevent cheating and ‘gaming’, so that society as a whole can trust such reported sustainability metrics,” added Co-PI Zadok.
A key contribution of the project is the design of metrics that holistically capture the sustainability costs of data center computing. The metrics include operational costs associated with clean and unclean energy usage in data centers, and embodied carbon costs involved in the manufacture, transport and disposal of data center equipment. Using such holistic metrics, the team is designing sustainable versions of computer services such as energy-efficient databases and machine learning frameworks.
The project also seeks to leverage energy-efficient “edge” nodes which have minimal compute capabilities but consume little power. By investigating the tradeoff between lower compute capabilities and low power consumption, the team aims to identify workloads and scenarios under which such edge nodes can provide more sustainable operations than commodity cloud servers.
The NSF CNS Large program funds “proposals tackling ambitious problems in computing and networking that are well suited to an integrated systems-oriented approach.’’ In the 2022 funding period, this SBU sustainability project was only one of three NSF CNS Large projects that were funded.
“This funding is an incredible testament to our capabilities in this space because the NSF CNS Large program is extremely competitive,” said Samir Das, chair of the Department of Computer Science. Congratulations to the entire team!”
Add comment