ABSTRACT
Cloud platforms' growing energy demand and carbon emissions are raising concern about their environmental sustainability. The current approach to enabling sustainable clouds focuses on improving energy-efficiency and purchasing carbon offsets. These approaches have limits: many cloud data centers already operate near peak efficiency, and carbon offsets cannot scale to near zero carbon where there is little carbon left to offset. Instead, enabling sustainable clouds will require applications to adapt to when and where unreliable low-carbon energy is available. Applications cannot do this today because their energy use and carbon emissions are not visible to them, as the energy system provides the rigid abstraction of a continuous, reliable energy supply. This vision paper instead advocates for a "carbon first" approach to cloud design that elevates carbon-efficiency to a firs--class metric. To do so, we argue that cloud platforms should virtualize the energy system by exposing visibility into, and software-defined control of, it to applications, enabling them to define their own abstractions for managing energy and carbon emissions based on their own requirements.
Supplemental Material
- 2018. OpenAI Blog, AI and Compute. https://openai.com/blog/ai-and-compute/.Google Scholar
- 2019. Reuters, Amazon Vows to be Carbon Neutral by 2040, buying 100,000 Electric Vans. https://www.reuters.com/article/us-amazon-environment/amazon-vows-to-be-carbon-neutral-by-2040-buying-100000-electric-vans-idUSKBN1W41ZV.Google Scholar
- 2020. Amazon EC2 Spot Instances. https://aws.amazon.com/ec2/spot/.Google Scholar
- 2020. Azure Spot Virtual Machines. https://azure.microsoft.com/en-us/pricing/spot/.Google Scholar
- 2020. Electricity Map. https://www.electricitymap.org/map.Google Scholar
- 2020. Google Preemptible Virtual Machines. https://cloud.google.com/preemptible-vms.Google Scholar
- 2021. Carbon free energy for Google Cloud regions. https://cloud.google.com/sustainability/region-carbon.Google Scholar
- 2021. Google Data Centers Efficiency. google.com/about/datacenters/efficiency/.Google Scholar
- 2021. Greenhouse Gas Protocol. https://ghgprotocol.org/.Google Scholar
- Nicola Acutt. 2018. Radius: Stories at the Edge, Achieving Carbon Neutrality. https://www.vmware.com/radius/achieving-carbon-neutrality/.Google Scholar
- A. Agarwal, J. Sun, S. Noghabi, S. Iyengar, A. Badam, R. Chandra, S. Seshan, and S. Kalyanaraman. 2021. Virtual Battery: Redesigning Cloud Computing for Renewable Energy. In HotNets.Google Scholar
- B. Alcott. [n.d.]. Jevons' Paradox. Ecological Economics 54, 1 ([n.d.]), 9--21.Google Scholar
- E. Bender, T. Gebru, A. McMillan-Major, and S. Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. In ACM FAccT.Google Scholar
- Mark Bohr. 2007. A 30 Year Retrospective on Dennard's MOSFET Scaling Paper. IEEE Solid-State Circuits Society Newsletter 12, 1 (Winter 2007), 11--13.Google ScholarCross Ref
- Greg Bothun. 2020. Basics of Solar Energy. http://zebu.uoregon.edu/disted/ph162/l4.html.Google Scholar
- Jeffrey S Chase, Darrell C Anderson, Prachi N Thakar, Amin M Vahdat, and Ronald P Doyle. 2001. Managing Energy and Server Resources in Hosting Centers. ACM SIGOPS Operating Systems Review 35, 5, 103--116.Google ScholarDigital Library
- A. Chien. 2021. Driving the Cloud to True Zero Carbon. CACM 64, 2 (February 2021).Google ScholarDigital Library
- Alyssa Daniels. 2020. Environmental Leader, Google Signs PPA for 140MW from Solar Farm in Texas. https://www.environmentalleader.com/2020/09/google-candela-texas-solar-ppa/.Google Scholar
- 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
- Dawson R. Engler, M. Frans Kaashoek, and James O'Toole. 1995. Exokernel: An Operating System Architecture for Application-Level Resource Management. In ACM Symposium on Operating System Principles (SOSP), Vol. 29. 251--266.Google ScholarDigital Library
- Darrell Etherington. 2020. TechCrunch, Google Claims Net Zero Carbon Footprint over its Entire Lifetime, Aims to only use Carbon-Free Energy by 2030. https://techcrunch.com/2020/09/14/google-claims-net-zero-carbon-footprint-over-its-entire-lifetime-aims-to-only-use-carbon-free-energy-by-2030/.Google Scholar
- S. Evans. 2020. CarbonBrief, Solar is now 'cheapest electricity in history', confirms IEA. https://www.carbonbrief.org/solar-is-now-cheapest-electricity-in-history-confirms-iea.Google Scholar
- Inigo Goiri, Ryan Beauchea, Kien Le, Thu D. Nguyen, Md. E. Haque, Jordi Guitart, Jordi Torres, and Ricardo Bianchini. 2011. GreenSlot: Scheduling Energy Consumption in Green Datacenters. In ACM/IEEE International Conference for High Performance Computing, Networking, Storage and Analysis (SC). Seattle, Washington, 449--471.Google ScholarDigital Library
- Inigo Goiri, William Katsak, Kien Le, Thu D. Nguyen, and Ricardo Bianchini. 2013. Parasol and GreenSwitch: Managing Datacenters Powered by Renewable Energy. In ACM Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), Vol. 48. Houston, Texas, 51--64.Google ScholarDigital Library
- I. Goiri, T. Nguyen, and R. Bianchini. 2015. CoolAir: Temperature- and Variation-Aware Management for Free-Cooled Datacenters. In ASPLOS, Vol. 50. 253--265.Google ScholarDigital Library
- Synergy Research Group. 2020. Hyperscale Data Center Count Reaches 541 in Mid-2020; Another 176 in the Pipeline. Technical Report. Synergy Research Group, Reno, NV (United States). https://www.srgresearch.com/articles/hyperscale-data-center-count-reaches-541-mid-2020-another-176-pipelineGoogle Scholar
- Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony Joseph, Randy Katz, Scott Shenker, and Ion Stoica. 2011. Mesos: A Platform for Fine-grained Resource Sharing in the Data Center. In USENIX Symposium on Networked Systems Design and Implementation (NSDI). Boston, Massachusetts, 295--308.Google Scholar
- Jeremy Hsu. 2019. How YouTube led to Google's cloud-gaming service: The tech that made YouTube work everywhere promises to do the same for games. IEEE Spectrum 56, 09 (2019), 9--10.Google ScholarDigital Library
- C. Imes, H. Zhang, K. Zhao, and H. Hoffman. 2019. CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping. In International Conference on Autonomic Computing (ICAC). 31--41.Google Scholar
- Rishikesh Jha, Stephen Lee, Srinivasan Iyengar, Mohammad H. Hajiesmaili, David Irwin, and Prashant Shenoy. 2020. Emission-aware Energy Storage Scheduling for a Greener Grid. In ACM International Conference on Future Energy Systems (e-Energy). 363--373.Google ScholarDigital Library
- Penny Jones. 2012. DataCenterDynamics, Apple Confirms Solar Farm at Maiden Data Center. https://www.datacenterdynamics.com/en/news/apple-conirms-solar-farm-at-maiden-data-center/.Google Scholar
- J. Koomey. 2011. Growth in Data Center Electricity Use 2005 to 2010. https://www.koomey.com/research.html.Google Scholar
- MLPerf. 2021. MLPerf™ v1.0 Inference Closed-Power ResNet-v1.5 server/offline (entries 1.0-{70,72,73,74}). https://mlcommons.org/en/inference-datacenter-10/Google Scholar
- E. Niiler. 2020. Wired, Do Carbon Offsets Really Work? It Depends on the Details. https://www.wired.com/story/do-carbon-offsets-really-work-it-depends-on-the-details/.Google Scholar
- Kevin O'Sullivan. 2020. The Irish Times, Facebook Commits to Net-Zero Carbon Emissions by 2030. https://www.irishtimes.com/news/environment/facebook-commits-to-net-zero-carbon-emissions-by-2030-1.4354701.Google Scholar
- Darshan S. Palasamudram, Ramesh K. Sitaraman, Bhuvan Urgaonkar, and Rahul Urgaonkar. 2012. Using Batteries to Reduce Power Costs of Internet-Scale Distributed Networks. In ACM Symposium on Cloud Computing (SoCC). 1--14.Google ScholarDigital Library
- Srinivas Pandruvada. 2014. Running Average Power Limit. https://01.org/blogs/2014/running-average-power-limit-%E2%80%93-rapl.Google Scholar
- S. Pelley, D. Meisner, P. Zandevakili, T. Wenisch, and J. Underwood. 2010. Power Routing: Dynamic Power Provisioning in the Data Center. In ASPLOS, Vol. 38. 231--242.Google Scholar
- A. Qureshi, R. Weber, H. Balakrishnan, J. Guttag, and B. Maggs. 2009. Cutting the Electric Bill for Internet-Scale Systems. In SIGCOMM. 123--134.Google Scholar
- A. Radovanovic. 2020. Google Blog, Our data centers now work harder when the sun shines and wind blows. https://blog.google/inside-google/infrastructure/data-centers-work-harder-sun-shines-wind-blows/.Google Scholar
- Jerome H. Saltzer, David P. Reed, and David D. Clark. 1984. End-To-End Arguments in System Design. ACM Transactions on Computer Systems 2, 4 (November 1984), 277--288.Google ScholarDigital Library
- Prateek Sharma, Tian Guo, Xin He, David Irwin, and Prashant Shenoy. 2016. Flint: Batch-Interactive Data-Intensive Processing for Transient Servers. In ACM European Conference on Computer Systems (EuroSys). London, United Kingdom, 1--15.Google ScholarDigital Library
- Prateek Sharma, Stephen Lee, Tian Guo, David Irwin, and Prashant Shenoy. 2015. SpotCheck: Designing a Derivative Cloud on the Spot Market. In ACM European Conference on Computer Systems (EuroSys). Bordeaux, France, 1--15.Google ScholarDigital Library
- Arman Shehabi, Sarah Josephine Smith, Dale A. Sartor, Richard E. Brown, Magnus Herrlin, Jonathan G. Koomey, Eric R. Masanet, Nathanial Horner, Ines Lima Azevedo, and William Linter. 2016. United States Data Center Energy Usage Report. Technical Report LBNL-1005775. Lawrence Berkeley National Lab (LBL).Google Scholar
- Kai Shen, Arrvindh Shriraman, Sandhya Dwarkadas, Xiao Zhang, and Zhuan Chen. 2013. Power Containers: An OS Facility for Fine-grained Power and Energy Management on Multicore Servers. In ACM Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS). 65--76.Google Scholar
- Prashant Shenoy and Thomas Wenisch. 2015. NSF Workshop on Sustainable Data Centers. Technical Report. National Science Foundation.Google Scholar
- Rahul Singh, David Irwin, Prashant Shenoy, and K.K. Ramakrishnan. 2013. Yank: Enabling Green Data Centers to Pull the Plug. In USENIX Symposium on Networked Systems Design and Implementation (NSDI). 143--156.Google Scholar
- Brad Smith. 2020. Official Microsoft Blog, Microsoft will be Carbon Negative by 2030. https://blogs.microsoft.com/blog/2020/01/16/microsoft-will-be-carbon-negative-by-2030/.Google Scholar
- S. Sorrell. 2009. Jevons' Paradox Revisited: The Evidence for Backfire from Improved Energy Efficiency. 37, 4 (2009), 1456--1469.Google Scholar
- Rick Stevens, Valerie Taylor, Jeff Nichols, Arthur Barney Maccabe, Katherine Yelick, and David Brown. 2020. AI for Science. Technical Report. Argonne National Lab.(ANL), Argonne, IL (United States).Google Scholar
- Emma Strubell, Ananya Ganesh, and Andrew McCallum. 2020. Energy and Policy Considerations for Modern Deep Learning Research. In AAAI Conference on Artificial Intelligence (AAAI). 13693--13696.Google Scholar
- Supreeth Subramanya, Tian Guo, Prateek Sharma, David Irwin, and Prashant Shenoy. 2015. SpotOn: A Batch Computing Service for the Spot Market. In Proceedings of the Sixth ACM Symposium on Cloud Computing (SoCC). Kohala Coast, Hawai'i, 1--13.Google ScholarDigital Library
- J. Switzer, R. McGuinness, P. Pannuto, G. Porter, A. Schulman, and B. Raghavan. 2021. TerraWatt: Sustaining Sustainable Computing of Containers in Containers. Technical Report. arXiv:2102.06614 https://arxiv.org/abs/2102.06614Google Scholar
- Domenico Talia. 2013. Clouds for scalable big data analytics. Computer 46, 5 (2013), 98--101.Google ScholarDigital Library
- Maud Texier. 2021. Google Cloud: A timely New Approach to Certifying Clean Energy. https://cloud.google.com/blog/topics/sustainability/t-eacs-offer-new-approach-to-certifying-clean-energy.Google Scholar
- R. Urgaonkar, B. Urgaonkar, M. Neely, and A. Sivasubramaniam. 2011. Optimal Power Cost Management Using Stored Energy in Data Centers. In SIGMETRICS. 221--232.Google Scholar
- M. Weiser, B. Welch, A. Demers, and S. Shenker. 1994. Scheduling for Reduced CPU Energy. In OSDI.Google Scholar
- J. Wilkes. 2020. Google Cluster-Usage Traces v3. Technical Report. Google Inc. Posted at https://github.com/google/cluster-data/blob/master/ClusterData2019.md.Google Scholar
- Ying Yan, Yanjie Gao, Yang Chen, Zhongxin Guo, Bole Chen, and Thomas Moscibroda. 2016. TR-Spark: Transient Computing for Big Data Analytics. In ACM Symposium on Cloud Computing (SoCC). Santa Clara, California, 484--496.Google ScholarDigital Library
- Youngseok Yang, Geon-Woo Kim, Won Wook Song, Yunseong Lee, Andrew Chung, Zhengping Qian, Brian Cho, and Byung-Gon Chun. 2017. Pado: A Data Processing Engine for Harnessing Transient Resources in Datacenters. In ACM European Conference on Computer Systems (EuroSys). Belgrade, Serbia, 575--588.Google ScholarDigital Library
- Matei Zaharia, Mosharaf Chowdhury, Tathagata Das, Ankur Dave, Justin Ma, Murphy McCauley, Michael Franklin, Scott Shenker, and Ion Stoica. 2012. Resilient Distributed Datasets: A Fault-Tolerant Abstraction for In-Memory Cluster Computing. In USENIX Symposium on Networked System Design and Implementation (NSDI). 15--28.Google Scholar
Index Terms
- Enabling Sustainable Clouds: The Case for Virtualizing the Energy System
Recommendations
Enabling Efficient Hypervisor-as-a-Service Clouds with Ephemeral Virtualization
VEE '16: Proceedings of the12th ACM SIGPLAN/SIGOPS International Conference on Virtual Execution EnvironmentsWhen considering a hypervisor, cloud providers must balance conflicting requirements for simple, secure code bases with more complex, feature-filled offerings. This paper introduces Dichotomy, a new two-layer cloud architecture in which the roles of the ...
Enabling Efficient Hypervisor-as-a-Service Clouds with Ephemeral Virtualization
VEE '16When considering a hypervisor, cloud providers must balance conflicting requirements for simple, secure code bases with more complex, feature-filled offerings. This paper introduces Dichotomy, a new two-layer cloud architecture in which the roles of the ...
Staring at Clouds
Cloud computing's premise is to lower computing costs by providing computational resources in a shared infrastructure. This could be a godsend to smaller organizations, but interoperability and security challenges still exist for this emerging ...
Comments