Does not compute: Avoiding pitfalls assessing the Internet's energy and carbon impacts
Status:: 🟩
Links:: Global energy consumption & carbon emissions of the whole ICT sector
Metadata
Authors:: Koomey, Jonathan; Masanet, Eric
Title:: Does not compute: Avoiding pitfalls assessing the Internet's energy and carbon impacts
Publication Title:: "Joule"
Date:: 2021
URL:: https://www.sciencedirect.com/science/article/pii/S2542435121002117
DOI:: 10.1016/j.joule.2021.05.007
Bibliography
Koomey, J., & Masanet, E. (2021). Does not compute: Avoiding pitfalls assessing the Internet’s energy and carbon impacts. Joule, 5(7), 1625–1628. https://doi.org/10.1016/j.joule.2021.05.007
Zotero
Type:: #zotero/journalArticle
Keywords:: [✅, 💎, Climate Footprint]
Relations
Abstract
Jonathan Koomey is president of Koomey Analytics and has in the past been a visiting professor at Stanford University, Yale University, and UC Berkeley. He's one of the leading international experts on the economics of climate solutions and the energy and environmental effects of information technology. Dr. Koomey holds M.S. and Ph.D. degrees from the Energy and Resources Group at UC Berkeley and an A.B. in History and Science from Harvard University. He is the author or coauthor of more than 200 articles and reports and nine books, including Turning Numbers into Knowledge: Mastering the Art of Problem Solving and Cold Cash, Cool Climate: Science-Based Advice for Ecological Entrepreneurs. More at http://www.koomey.com. Eric Masanet is the Mellichamp Chair in Sustainability Science for Emerging Technologies at the University of California, Santa Barbara, where he holds appointments in the Bren School of Environmental Science and Management and the Department of Mechanical Engineering. He has authored more than 130 scientific publications on sustainability modeling of energy and materials demand systems, with particular focuses on data centers and IT systems. He holds a Ph.D. in mechanical engineering from UC Berkeley, with a focus on sustainable manufacturing.
Notes & Annotations
Keynote talk by Koomey at the iTherm 2021 technical conference:
https://href.li/?https://www.mediafire.com/file/79h3mx3rg1jmkyl/KoomeyslidesforiTherm210601-v5.pdf/file
📑 Annotations (imported on 2023-04-04#18:21:07)
The past decade saw highly cited topdown projections of data-center energy use that either underestimated subsequent energy-efficiency gains or ignored them altogether, leading to predictions of massive energy-demand growth by the decade’s end.
A second pitfall is assuming that shortterm changes in computing services must lead to proportional and immediate changes in electricity use.
A third pitfall is making long-term projections, even if these projections explicitly account for technological change in efficiency and other key drivers of electricity use. IT changes so quickly that even projections extending only a few years are highly uncertain (but can sometimes be valid and useful). Applying exponential growth rates in demand growth for more than a few years can result in eye-popping projected changes
A fourth pitfall is drawing broad conclusions based on trends in only one part of the IT system. Analytical rigor comes from analyzing a whole system, which is often difficult due to pervasive data gaps. However, just focusing on one highly visible part of the system can give a mistaken impression about what’s happening with the electricity used by the whole system.
For example, the tremendous growth in the cloud data-center segment (which includes the world’s largest ‘‘hyperscale’’ data centers) has led some to predict massive future growth in global data-center energy use. Between 2010 and 2018, the workloads hosted by this segment increased by 2,600%, whereas its estimated electricity use increased by 500%. Despite the rapid growth of this segment, the global energy use of all data centers grew far more modestly, rising by less than 10%.
Much of the compute output from cloud/hyperscale data centers has displaced traditional in-house data centers that use several times more electricity to perform the same tasks.