ESnet Internship Leads to Book Chapter on Energy Efficiency for Datacenters, Networks

Just as research and education networks have expanded the capabilities of scientists, a 2011 internship at ESnet led Baris Aksanli, then a Ph.D. student in computer science and engineering from the University of California San Diego, to broaden the scope of his research in tapping renewable energy to reduce the carbon footprint of data centers and networks. The authors found that centers can increase their use of renewable energy while also reducing costs.

The result is a chapter in a new book, Computational Sustainability, edited by  Joerg Laessig, Kristian Kersting and Katharina Morik and  published on May 30, 2016 by Springer. Aksanli, along with Jagannathan Venkatesh and Tajana Simunic Rosing of UC San Diego and ESnet’s Inder Monga collaborated on the chapter on “Renewable Energy Prediction for Improved Utilization and Efficiency in Datacenters and Backbone Networks.”

Since submitting the chapter a couple of years ago, Aksanli has graduated and will start working at San Diego State University as an assistant professor in August 2016. His research interests include energy efficient cyber physical systems, human behavior modeling in the Internet of Things and big data for energy-efficient embedded systems.

In their abstract, the authors write that “Datacenters are one of the important global energy consumers and carbon producers. However, their tight service level requirements prevent easy integration with highly variable renewable energy sources.”

Aksanli said his personal interest in energy efficiency led to studying how a single datacenter could become more energy efficient, but after working at Berkeley Lab and ESnet, the natural extension was to look at a group of networked datacenters, such as Department of Energy computing centers connected via ESnet.

But because of the critical role high performance computer centers play in DOE research, it’s difficult to plug them into highly variable renewable energy sources. But using prediction to determine short-term green energy availability can mitigate this vulnerability. To reduce brown energy consumption costs, the authors first looked at existing short-term solar and wind energy prediction methods, then leveraged prediction to allocate and mitigate workloads across geographically distributed datacenters.

“The results show that prediction enables up to 90 percent green energy utilization, a 3x improvement over the existing methods. The cost minimization algorithm reduces expenses by up to 16 percent and increases performance by 27 percent when migrating workloads across datacenters,” they wrote in their abstract. “ Furthermore, the savings increase up to 30 percent  compared with no migration when servers are made energy-proportional.”