ABSTRACT
The advent of cloud computing has provided people around the world with unprecedented access to computational power and enabled rapid growth in technologies such as machine learning, the computational demands of which incur a high energy cost and a commensurate carbon footprint. As a result, recent scholarship has called for better estimates of the greenhouse gas impact of AI: data scientists today do not have easy or reliable access to measurements of this information, which precludes development of actionable tactics. We argue that cloud providers presenting information about software carbon intensity to users is a fundamental stepping stone towards minimizing emissions.
In this paper, we provide a framework for measuring software carbon intensity, and propose to measure operational carbon emissions by using location-based and time-specific marginal emissions data per energy unit. We provide measurements of operational software carbon intensity for a set of modern models covering natural language processing and computer vision applications, and a wide range of model sizes, including pretraining of a 6.1 billion parameter language model. We then evaluate a suite of approaches for reducing emissions on the Microsoft Azure cloud compute platform: using cloud instances in different geographic regions, using cloud instances at different times of day, and dynamically pausing cloud instances when the marginal carbon intensity is above a certain threshold. We confirm previous results that the geographic region of the data center plays a significant role in the carbon intensity for a given cloud instance, and find that choosing an appropriate region can have the largest operational emissions reduction impact. We also present new results showing that the time of day has meaningful impact on operational software carbon intensity.Finally, we conclude with recommendations for how machine learning practitioners can use software carbon intensity information to reduce environmental impact.
- Lasse F. Wolff Anthony, Benjamin Kanding, and Raghavendra Selvan. 2020. Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models. arxiv:2007.03051 [cs.CY]Google Scholar
- Rhonda Ascierto and A Lawrence. 2020. Uptime institute global data center survey 2020. Uptime Institute 2(2020).Google Scholar
- Nesrine Bannour, Sahar Ghannay, Aurélie Névéol, and Anne-Laure Ligozat. 2021. Evaluating the carbon footprint of NLP methods: a survey and analysis of existing tools. In EMNLP, Workshop SustaiNLP.Google ScholarCross Ref
- Abeba Birhane, Pratyusha Kalluri, Dallas Card, William Agnew, Ravit Dotan, and Michelle Bao. 2021. The values encoded in machine learning research. arXiv preprint arXiv:2106.15590(2021).Google Scholar
- Buildcomputers.net. 2021. Power Consumption of PC Components in Watts. https://www.buildcomputers.net/power-consumption-of-pc-components.htmlGoogle Scholar
- Jacques A de Chalendar and Sally M Benson. 2019. Why 100% renewable energy is not enough. Joule 3, 6 (2019), 1389–1393.Google ScholarCross Ref
- Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. 2009. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition. Ieee, 248–255.Google ScholarCross Ref
- Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. arxiv:1810.04805 [cs.CL]Google Scholar
- Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929(2020).Google Scholar
- Jim Gao. 2014. Machine learning applications for data center optimization. (2014).Google Scholar
- Michael Gillenwater. 2008. Redefining RECs—Part 1: untangling attributes and offsets. Energy Policy 36, 6 (2008), 2109–2119.Google ScholarCross Ref
- Google. 2021. Carbon free energy for Google Cloud regions. https://cloud.google.com/sustainability/region-carbonGoogle Scholar
- Google. 2021. Helping you pick the greenest region for your Google Cloud resources. https://cloud.google.com/blog/topics/sustainability/pick-the-google-cloud-region-with-the-lowest-co2Google Scholar
- Abhishek Gupta, Camylle Lanteigne, and Sara Kingsley. 2020. SECure: A Social and Environmental Certificate for AI Systems. arXiv preprint arXiv:2006.06217(2020).Google Scholar
- Udit Gupta, Young Geun Kim, Sylvia Lee, Jordan Tse, Hsien-Hsin S Lee, Gu-Yeon Wei, David Brooks, and Carole-Jean Wu. 2021. Chasing Carbon: The Elusive Environmental Footprint of Computing. In 2021 IEEE International Symposium on High-Performance Computer Architecture (HPCA). IEEE, 854–867.Google Scholar
- K. Hazelwood, S. Bird, D. Brooks, S. Chintala, U. Diril, D. Dzhulgakov, M. Fawzy, B. Jia, Y. Jia, A. Kalro, J. Law, K. Lee, J. Lu, P. Noordhuis, M. Smelyanskiy, L. Xiong, and X. Wang. 2018. Applied Machine Learning at Facebook: A Datacenter Infrastructure Perspective. In 2018 IEEE International Symposium on High Performance Computer Architecture (HPCA). 620–629. https://doi.org/10.1109/HPCA.2018.00059Google ScholarCross Ref
- Kees Hertogh. 2021. Empowering cloud sustainability with the Microsoft Emissions Impact Dashboard. https://azure.microsoft.com/en-us/blog/empowering-cloud-sustainability-with-the-microsoft-emissions-impact-dashboard/Google Scholar
- Zeshan Hyder, Keng Siau, and Fiona Nah. 2019. Artificial intelligence, machine learning, and autonomous technologies in mining industry. Journal of Database Management (JDM) 30, 2 (2019), 67–79.Google ScholarCross Ref
- Forrest Iandola, Matt Moskewicz, Sergey Karayev, Ross Girshick, Trevor Darrell, and Kurt Keutzer. 2014. Densenet: Implementing efficient convnet descriptor pyramids. arXiv preprint arXiv:1404.1869(2014).Google Scholar
- International Energy Authority (IEA). 2020. Energy Technology Perspectives 2020. https://www.iea.org/reports/energy-technology-perspectives-2020Google Scholar
- Lynn Kaack, Priya Donti, Emma Strubell, George Kamiya, Felix Creutzig, and David Rolnick. 2021. Aligning artificial intelligence with climate change mitigation. (2021).Google Scholar
- Young Geun Kim and Carole-Jean Wu. 2021. AutoFL: Enabling Heterogeneity-Aware Energy Efficient Federated Learning. In MICRO-54: 54th Annual IEEE/ACM International Symposium on Microarchitecture. 183–198.Google Scholar
- Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dandres. 2019. Quantifying the carbon emissions of machine learning. arXiv preprint arXiv:1910.09700(2019).Google Scholar
- Nevena Lazic, Tyler Lu, Craig Boutilier, MK Ryu, Eehern Jay Wong, Binz Roy, and Greg Imwalle. 2018. Data center cooling using model-predictive control. (2018).Google Scholar
- Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. 1998. Gradient-based learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278–2324.Google ScholarCross Ref
- Anne-Laure Ligozat, Julien Lefèvre, Aurélie Bugeau, and Jacques Combaz. 2021. Unraveling the hidden environmental impacts of AI solutions for environment. arXiv preprint arXiv:2110.11822(2021).Google Scholar
- Eric Masanet, Arman Shehabi, Nuoa Lei, Sarah Smith, and Jonathan Koomey. 2020. Recalibrating global data center energy-use estimates. Science 367, 6481 (2020), 984–986.Google Scholar
- Valérie Masson-Delmotte, Panmao Zhai, Hans-Otto Pörtner, Debra Roberts, Jim Skea, Priyadarshi R Shukla, Anna Pirani, Wilfran Moufouma-Okia, Clotilde Péan, Roz Pidcock, 2018. Global warming of 1.5 C. An IPCC Special Report on the impacts of global warming of 1, 5(2018).Google Scholar
- Microsoft Azure. 2021. Azure sustainability. https://azure.microsoft.com/en-us/global-infrastructure/sustainability/#overviewGoogle Scholar
- Vasilii Mosin, Roberto Aguilar, Alexander Platonov, Albert Vasiliev, Alexander Kedrov, and Anton Ivanov. 2019. Remote sensing and machine learning for tree detection and classification in forestry applications. In Image and Signal Processing for Remote Sensing XXV, Vol. 11155. International Society for Optics and Photonics, 111550F.Google Scholar
- NeurIPS 2021 Conference. 2021. NeurIPS 2021 Paper Checklist Guidelines. https://neurips.cc/Conferences/2021/PaperInformation/PaperChecklistGoogle Scholar
- Vito Alexander Nordloh, Anna Roubícková, and Nick Brown. 2020. Machine Learning for Gas and Oil Exploration. arXiv preprint arXiv:2010.04186(2020).Google Scholar
- David Patterson, Joseph Gonzalez, Urs Hölzle, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean. 2022. The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink. TexRxiv (2022).Google Scholar
- David Patterson, Joseph Gonzalez, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean. 2021. Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350(2021).Google Scholar
- Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. arXiv preprint arXiv:1910.10683(2019).Google Scholar
- Joeri Rogelj, Oliver Geden, Annette Cowie, and Andy Reisinger. 2021. Net-zero emissions targets are vague: three ways to fix. Nature 591(2021), 365–368.Google ScholarCross Ref
- Victor Sanh, Albert Webson, Colin Raffel, Stephen H Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, 2021. Multitask prompted training enables zero-shot task generalization. arXiv preprint arXiv:2110.08207(2021).Google Scholar
- Victor Schmidt, Kamal Goyal, Aditya Joshi, Boris Feld, Liam Conell, Nikolas Laskaris, Doug Blank, Jonathan Wilson, Sorelle Friedler, and Sasha Luccioni. 2021. CodeCarbon: Estimate and Track Carbon Emissions from Machine Learning Computing.Google Scholar
- Roy Schwartz, Jesse Dodge, Noah A Smith, and Oren Etzioni. 2020. Green AI. Commun. ACM 63, 12 (2020), 54–63.Google ScholarDigital Library
- Amazon Web Services. 2021. Sustainability in the Cloud. https://sustainability.aboutamazon.com/environment/the-cloudGoogle Scholar
- Omar Shaikh, Jon Saad-Falcon, Austin P Wright, Nilaksh Das, Scott Freitas, Omar Asensio, and Duen Horng Chau. 2021. EnergyVis: Interactively Tracking and Exploring Energy Consumption for ML Models. Association for Computing Machinery, New York, NY, USA. https://doi.org/10.1145/3411763.3451780Google ScholarDigital Library
- Emma Strubell, Ananya Ganesh, and Andrew McCallum. 2019. Energy and policy considerations for deep learning in NLP. arXiv preprint arXiv:1906.02243(2019).Google Scholar
- Neil C Thompson, Kristjan Greenewald, Keeheon Lee, and Gabriel F Manso. 2020. The computational limits of deep learning. arXiv preprint arXiv:2007.05558(2020).Google Scholar
- Romal Thoppilan, Daniel De Freitas, Jamie Hall, Noam Shazeer, Apoorv Kulshreshtha, Heng-Tze Cheng, Alicia Jin, Taylor Bos, Leslie Baker, Yu Du, YaGuang Li, Hongrae Lee, Huaixiu Steven Zheng, Amin Ghafouri, Marcelo Menegali, Yanping Huang, Maxim Krikun, Dmitry Lepikhin, James Qin, Dehao Chen, Yuanzhong Xu, Zhifeng Chen, Adam Roberts, Maarten Bosma, Yanqi Zhou, Chung-Ching Chang, Igor Krivokon, Will Rusch, Marc Pickett, Kathleen Meier-Hellstern, Meredith Ringel Morris, Tulsee Doshi, Renelito Delos Santos, Toju Duke, Johnny Soraker, Ben Zevenbergen, Vinodkumar Prabhakaran, Mark Diaz, Ben Hutchinson, Kristen Olson, Alejandra Molina, Erin Hoffman-John, Josh Lee, Lora Aroyo, Ravi Rajakumar, Alena Butryna, Matthew Lamm, Viktoriya Kuzmina, Joe Fenton, Aaron Cohen, Rachel Bernstein, Ray Kurzweil, Blaise Aguera-Arcas, Claire Cui, Marian Croak, Ed Chi, and Quoc Le. 2022. LaMDA: Language Models for Dialog Applications. arxiv:2201.08239 [cs.CL]Google Scholar
- Georgina Torbet. 2019. How Much Energy Does Your PC Use? (And 8 Ways to Cut It Down). https://www.makeuseof.com/tag/much-energy-pc-use-8-ways-cut/Google Scholar
- United States Environmental Protection Agency. 2021. Greenhouse Gas Equivalencies Calculator. https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculatorGoogle Scholar
- US Department of Energy. 2021. Energy-Efficient Cooling Control Systems for Data Centers. https://www.energy.gov/eere/amo/energy-efficient-cooling-control-systems-data-centersGoogle Scholar
- Adina Williams, Nikita Nangia, and Samuel R Bowman. 2017. A broad-coverage challenge corpus for sentence understanding through inference. arXiv preprint arXiv:1704.05426(2017).Google Scholar
- Carole-Jean Wu, Ramya Raghavendra, Udit Gupta, Bilge Acun, Newsha Ardalani, Kiwan Maeng, Gloria Chang, Fiona Aga Behram, James Huang, Charles Bai, 2021. Sustainable ai: Environmental implications, challenges and opportunities. arXiv preprint arXiv:2111.00364(2021).Google Scholar
Index Terms
- Measuring the Carbon Intensity of AI in Cloud Instances
Recommendations
The Carbon Intensity Evaluation of Coal-fired Power Plants
IEEA '18: Proceedings of the 7th International Conference on Informatics, Environment, Energy and ApplicationsThe global carbon dioxide emissions gradually increase in the last few decades, among which the coal-fired power plants play magnificent role. Hence the carbon balance theory is in utilization to establish the carbon emission model for combustion and ...
Hybrid simulation and energy market based optimization of cement plants
This paper presents an approach for equipping a cement plant with a wind power plant, a battery storage and an optimized control in order to reduce electricity supply cost and carbon dioxide (CO2) emissions as cement manufacturing is a traditional ...
Reduction of carbon dioxide emissions by mineral carbonation
EE'10: Proceedings of the 5th IASME/WSEAS international conference on Energy & environmentThe study investigates the technologies that have the potential to provide feasible reduction of carbon dioxide (CO2) from a reference power plant. Particular focus has been given to mineral carbonation (at 1 bar) in which magnesium (Mg) and/or calcium (...
Comments