Measuring the Carbon Intensity of AI in Cloud Instances

Jesse Dodge; Taylor Prewitt; Remi Tachet Des Combes; Erika Odmark; Roy Schwartz; Emma Strubell; Alexandra Sasha Luccioni; Noah A. Smith; Nicole Decario; Will Buchanan

doi:10.1145/3531146.3533234

Measuring the Carbon Intensity of AI in Cloud Instances

Jesse Dodge, Taylor Prewitt, Remi Tachet Des Combes, Erika Odmark, Roy Schwartz, Emma Strubell, Alexandra Sasha Luccioni, Noah A. Smith, Nicole Decario, Will Buchanan

The Rachel and Selim Benin School of Engineering and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

99 Scopus citations

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.

Original language	English
Title of host publication	Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022
Publisher	Association for Computing Machinery
Pages	1877-1894
Number of pages	18
ISBN (Electronic)	9781450393522
DOIs	https://doi.org/10.1145/3531146.3533234
State	Published - 21 Jun 2022
Event	5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022 - Virtual, Online, Korea, Republic of Duration: 21 Jun 2022 → 24 Jun 2022

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022
Country/Territory	Korea, Republic of
City	Virtual, Online
Period	21/06/22 → 24/06/22

Bibliographical note

Publisher Copyright:
© 2022 Owner/Author.

Keywords

CO2
carbon awareness
carbon intensity
cloud
emissions
grid

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1145/3531146.3533234

Cite this

Dodge, J., Prewitt, T., Tachet Des Combes, R., Odmark, E., Schwartz, R., Strubell, E., Luccioni, A. S., Smith, N. A., Decario, N., & Buchanan, W. (2022). Measuring the Carbon Intensity of AI in Cloud Instances. In Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022 (pp. 1877-1894). (ACM International Conference Proceeding Series). Association for Computing Machinery. https://doi.org/10.1145/3531146.3533234

@inproceedings{987a751776564034976683c11e9ea85b,

title = "Measuring the Carbon Intensity of AI in Cloud Instances",

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.",

keywords = "CO2, carbon awareness, carbon intensity, cloud, emissions, grid",

author = "Jesse Dodge and Taylor Prewitt and {Tachet Des Combes}, Remi and Erika Odmark and Roy Schwartz and Emma Strubell and Luccioni, {Alexandra Sasha} and Smith, {Noah A.} and Nicole Decario and Will Buchanan",

note = "Publisher Copyright: {\textcopyright} 2022 Owner/Author.; 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022 ; Conference date: 21-06-2022 Through 24-06-2022",

year = "2022",

month = jun,

day = "21",

doi = "10.1145/3531146.3533234",

language = "אנגלית",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "1877--1894",

booktitle = "Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022",

}

Dodge, J, Prewitt, T, Tachet Des Combes, R, Odmark, E, Schwartz, R, Strubell, E, Luccioni, AS, Smith, NA, Decario, N & Buchanan, W 2022, Measuring the Carbon Intensity of AI in Cloud Instances. in Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 1877-1894, 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022, Virtual, Online, Korea, Republic of, 21/06/22. https://doi.org/10.1145/3531146.3533234

Measuring the Carbon Intensity of AI in Cloud Instances. / Dodge, Jesse; Prewitt, Taylor; Tachet Des Combes, Remi et al.
Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022. Association for Computing Machinery, 2022. p. 1877-1894 (ACM International Conference Proceeding Series).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Measuring the Carbon Intensity of AI in Cloud Instances

AU - Dodge, Jesse

AU - Prewitt, Taylor

AU - Tachet Des Combes, Remi

AU - Odmark, Erika

AU - Schwartz, Roy

AU - Strubell, Emma

AU - Luccioni, Alexandra Sasha

AU - Smith, Noah A.

AU - Decario, Nicole

AU - Buchanan, Will

PY - 2022/6/21

Y1 - 2022/6/21

N2 - 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.

AB - 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.

KW - CO2

KW - carbon awareness

KW - carbon intensity

KW - cloud

KW - emissions

KW - grid

UR - http://www.scopus.com/inward/record.url?scp=85133031151&partnerID=8YFLogxK

U2 - 10.1145/3531146.3533234

DO - 10.1145/3531146.3533234

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85133031151

T3 - ACM International Conference Proceeding Series

SP - 1877

EP - 1894

BT - Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022

PB - Association for Computing Machinery

T2 - 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022

Y2 - 21 June 2022 through 24 June 2022

ER -

Dodge J, Prewitt T, Tachet Des Combes R, Odmark E, Schwartz R, Strubell E et al. Measuring the Carbon Intensity of AI in Cloud Instances. In Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022. Association for Computing Machinery. 2022. p. 1877-1894. (ACM International Conference Proceeding Series). doi: 10.1145/3531146.3533234

Measuring the Carbon Intensity of AI in Cloud Instances

Abstract

Publication series

Conference

Bibliographical note

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this