Towards transparent and trustworthy cloud carbon accounting

Climate Change is arguably the biggest challenge that humanity faces today. Multiple trends such as the exponential explosion of data transfer, the emergence and popularity of power intensive workloads such as AI, and the flattening of Moore's law contribute to a rising concern over the increasing carbon footprint cost of digital computation. Any effective strategy to reduce the energy consumption and associated carbon footprint of computations must begin with an accurate and transparent quantification method. However, while most businesses today run a significant portion of their workloads on third party cloud environments, transparent carbon quantification of tenant workloads in cloud environments is lacking. This regretful situation inhibits reliable reporting of Scope 3 Green House Gas (GHG) by cloud users, meaningful comparison of cloud carbon efficiencies, and measurable reduction strategies. In this extended abstract we explain the unique challenges that arise in multi-tenant cloud environments, and propose and discuss an approach, consistent with the GHG Protocol, for cloud carbon footprint quantification. The quantification is a first step towards sustainable cloud environments, that employ dynamic controllers to quantify and reduce the carbon footprint at every layer of the cloud stack.

[1]  Sherif Abdelwahed,et al.  Power-Aware Modeling and Autonomic Management Framework for Distributed Computing Systems , 2012, Handbook of Energy-Aware and Green Computing.

[2]  Anders S. G. Andrae,et al.  On Global Electricity Usage of Communication Technology: Trends to 2030 , 2015 .

[3]  David Buchaca Prats,et al.  Proactive Container Auto-scaling for Cloud Native Machine Learning Services , 2020, 2020 IEEE 13th International Conference on Cloud Computing (CLOUD).

[4]  Romain Rouvoy,et al.  SelfWatts: On-the-fly Selection of Performance Events to Optimize Software-defined Power Meters , 2021, 2021 IEEE/ACM 21st International Symposium on Cluster, Cloud and Internet Computing (CCGrid).

[5]  Josep Lluís Berral,et al.  AI4DL: Mining Behaviors of Deep Learning Workloads for Resource Management , 2020, HotCloud.

[6]  Sherif Abdelwahed,et al.  Power-Aware Modeling and Autonomic Management Framework for Distributed Computing Systems , 2012 .

[7]  David Atienza,et al.  Containergy—A Container-Based Energy and Performance Profiling Tool for Next Generation Workloads , 2020 .

[8]  Ehsan Arianyan,et al.  An Energy-Aware Dynamic Resource Management Technique Using Deep Q-Learning Algorithm and Joint VM and Container Consolidation Approach for Green Computing in Cloud Data Centers , 2020, DCAI.

[9]  Hai Jin,et al.  Virtual Machine Power Accounting with Shapley Value , 2017, 2017 IEEE 37th International Conference on Distributed Computing Systems (ICDCS).

[10]  N. Jones.  How to stop data centres from gobbling up the world’s electricity , 2018, Nature.

[11]  David Mytton.  Hiding greenhouse gas emissions in the cloud , 2020, Nature Climate Change.

[12]  Simon Fong,et al.  An adaptive workload-aware power consumption measuring method for servers in cloud data centers , 2020, Computing.

[13]  Yonggang Wen,et al.  Transforming Cooling Optimization for Green Data Center via Deep Reinforcement Learning , 2017, IEEE Transactions on Cybernetics.

[14]  Oren Etzioni,et al.  Green AI , 2019, Commun. ACM.

[15]  René Pawlitzek,et al.  New universal sustainability metrics to assess edge intelligence , 2021, Sustain. Comput. Informatics Syst..

[16]  Dawei Li,et al.  An energy-efficient algorithm for virtual machine placement optimization in cloud data centers , 2020, Cluster Computing.

[17]  Dimos Poulikakos,et al.  Aquasar: A hot water cooled data center with direct energy reuse , 2012 .

[18]  Andrew McCallum,et al.  Energy and Policy Considerations for Deep Learning in NLP , 2019, ACL.

[19]  Bin Xia,et al.  Estimating Power Consumption of Containers and Virtual Machines in Data Centers , 2020, 2020 IEEE International Conference on Cluster Computing (CLUSTER).

[20]  Zong Woo Geem,et al.  Optimized Energy Cost and Carbon Emission-Aware Virtual Machine Allocation in Sustainable Data Centers , 2020, Sustainability.

[21]  L. Carloni.  Scalable Open-Source System-on-Chip Design , 2020 .

[22]  Romain Rouvoy,et al.  SmartWatts: Self-Calibrating Software-Defined Power Meter for Containers , 2020, 2020 20th IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGRID).