WTM20 - Sustainable ICT and ICT for Sustainability: a matter of limited resources by Ana Carolina Riekstin

Believing or not in climate change, being sustainable is ultimately a matter of limited resources: money, energy, emissions. We now talk about power budgets, carbon emissions, how to reach more people, and do more with less. Come to see what we as researchers and developers have to do with this.

Ana Carolina Riekstin
Montreal, Qc, Canada / Bell
Software Developer

Ana Carolina is a Brazilian based in Montreal who has been around research and development related to computer networks and network management, smart homes and cities, and sustainability. Currently a Software Developer at Bell, holds a PhD in computer engineering. Loves to talk, specially about her dog, Caco, and Brazilian cheese bread recipes. Still searching for the answer to life, the universe and everything.

Transcript

1. Sustainable ICT and ICT for Sustainability: a matter of limited resources Ana Carolina Riekstin
2. WHAT WHY WHO WHEN WHERE HOW
3. Sustainability Sustainability refers to a process or state that can be maintained at a certain level for as long as it is wanted Sustainable development: “development that meets the needs of the present without compromising the ability of future generations to meet their own needs” (Brundtland Report, Our Common Future) Information and Communications Technologies (ICT): Sustainable ICT ICT for Sustainability
4. International Energy Agency, 2017: https://www.iea.org/digital/
5. The ICT industry could be using 20% of all electricity in the world and emitting around 5.5% of the world’s carbon emissions by 2025 Anders Andrae, 2018 apud https://www.theguardian.com/environment/2017/dec/11/tsunami-of-data-could-consume-fifth-global-electricity-by-2025
6. Clicking Clean (Greenpeace), 2017: https://www.greenpeace.org/usa/global-warming/click-clean/
7. 20% reduction of global emissions + 2.5 billion extra people to the “knowledge economy” by 2030 → Health, education, and agriculture, manufacturing, smart buildings, energy etc ICT for Sustainability GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf
9. Infrastructures Software Development Applications ----- Metrics ----- --- Methods / Best Practices ---
10. Infra
11. Main components of electricity consumption for the IT sector Clicking Clean (Greenpeace), 2017: https://www.greenpeace.org/usa/global-warming/click-clean/
12. Riekstin, A.C et al. 2018: A Survey on Metrics and Measurement Tools for Sustainable Distributed Cloud Networks, IEEE Communications Surveys & Tutorials Carbon Intensity → PUE → GEC → ECR →
13. Riekstin, A.C et al. 2018: A Survey on Metrics and Measurement Tools for Sustainable Distributed Cloud Networks, IEEE Communications Surveys & Tutorials
14. Google Sustainability Report 2019: https://sustainability.google/reports/environmental-report-2019/
15. Facebook PUE/WUE dashboards: https://engineering.fb.com/data-center-engineering/open-sourcing-pue-wue-dashboards/
16. D. Dudkowski and K. Samdanis, 2012: Energy consumption monitoring techniques in communication networks, IEEE CCNC
17. For GHG emissions metrics Energy (Wh) Emission Factor (gCO2e/Wh) -- (gCO2e)
18. https://www.electricitymap.org/ Emission factors
19. IEEE-SA
20. Different methods/tools to use •Offloading tasks to reduce latency and energy (e.g. to the mobile-edge computing with data compression for transmission) Offloading •Set CPU/GPU timing settings to run at a lower clock rate Underclocking •Change clock level or power level (ACPI - Advanced Configuration and Power Interface) •Reduce the link rate (e.g., from 1Gbps to 100Mbps) (e.g. Adaptive Link Rate) Rating •Put the device in idle mode when there is no data to be sent in a way that allows it to wake up quickly when a new packet arrives (IEEE 802.3az - Energy Efficient Ethernet) Sleeping •To create more opportunities to put devices to sleep or in “reduced” mode •Or to send traffic or compute stuff where the energy is cleaner Traffic / Load Management Riekstin, A.C 2015: Orchestration of energy efficiency capabilities for a sustainable network management, Thesis
21. Example: PBNM + orchestration
22. Software
23. Some additional motivation… Energy efficiency can be a requirement • “turn-by-turn guided navigation should not drain more battery than the car can charge” • “under normal usage, a device with an XWh battery should last for Y hours” (Manotas et al. 2016) Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Computing Surveys
24. Bozzelli et al. 2014: A systematic literature review on green software metrics, VU Technical Report
25. To derive the energy consumption… • indirect energy measurements through models or performance counters • direct measurements, through hardware energy analyzers and sensors Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Computing Surveys
26. https://developer.android.com/studio/profile/energy-profiler
27. https://mlco2.github.io/impact/
28. Design patterns • The observer and decorator as the most energy-greedy design patterns Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Computing Surveys
29. Programming languages Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Computing Surveys Languages
30. Different methods/tools to use •Playing with threads can help identifying the best combination to reduce consumption and execution time; energy consumption increases as the number of threads increases until the number of threads reaches the number of CPU cores; after, the energy is reduced Parallelism •Approximate computing is an approach for sacrificing computation accuracy – when possible - to increase runtime performance or energy savings •One example is Memoization: store expensive function call results Approximate computing •Testing process that focuses on revealing defects and vulnerabilities in a computer program before its deployment phase can also help checking hot-spots regarding energy consumption Source code analysis Kambadur and Kim, 2014: An experimental survey of energy management across the stack, ACM OOPSLA’14
31. Tools / Best practices • https://wiki.cs.vu.nl/green_softwar e/Best_practices_for_energy_effici ent_software • Some examples studied by Kambadur & Kim: Kambadur and Kim, 2014: An experimental survey of energy management across the stack, ACM OOPSLA’14 https://wiki.cs.vu.nl/green_software/Best_practices_for_energy_efficient_software
32. Applications
33. United Nations
34. Buildings • “ICT → increased comfort and reduced energy and water bills” • Smart building solutions could cut 2.0Gt CO2e from the housing sector, reducing energy costs by $0.4 trillion and creating revenue opportunities of $0.4 trillion GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf
35. https://www.microsoft.com/en-us/itshowcase/microsoft-uses-machine-learning-to-develop-smart-energy-solutions x 43 buildings targeted for HVAC scheduling prediction in the Puget Sound area = savings will exceed $500,000 annually Saving more than 52,000 person- hours of reported discomfort due to low or high temperatures
36. DIY! • Get that emission factors we mentioned a couple of slides ago • Program your appliances to work at the best time of the day  “best time” can be even predicted using some machine learning techniques Riekstin et al., 2019: Time Series-Based GHG Emissions Prediction for Smart Homes, Transactions on Sustainable Computing
37. Commuting • ”ICT → help everyone reach their destinations faster, cheaper and safer” • Real-time traffic information, smart logistics, intelligent lighting and other ICT enabled solutions could abate 3.6Gt CO2e, including abatement from avoided travel GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf
38. Or working remotely • ”ICT-enabled telecommuting and virtual conferencing can save employees time and money” • E-Work could add $0.5 trillion of revenues while freeing up 100 hours per E-Worker annually GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf
39. And more • Energy: integration of renewables onto the grid, Smart Grids • Health: “a doctor in your pocket”, E-Health services to reach more people • Learning: education more accessible, engaging, flexible and affordable, may help to raise incomes • Manufacturing: more efficient use of resources, Smart manufacturing • Agriculture: Smart Agriculture will boost yields by 30%, avoid 20% of food waste and could deliver economic benefits worth $1.9 trillion; could also reduce water needs by 250 trillion liters and abate 2.0Gt CO2e GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf
40. And more! • “Girls fleeing FGM come from very remote villages, which until recently, were not on any map. There are no road signs and it was very difficult to reach the villages, particularly at night” • Volunteer, start mapping! https://tasks.hotosm.org/contribute?difficulty=ALL&organisation=Tanzania%20Development%20Trust
42. • Increasing traffic, devices, processing power → increased demand for energy, more GHG emissions • The need to be energy efficient, clients establishing power budgets, government asking for GHG accounting • Challenges and opportunities to improve the infrastructure, the software, and the applications  Metrics, methods, best practices... • And there is more to explore!  Machine learning methods and algorithms (e.g. Shoaib et al., 2015)  Reduce electronic waste and embodied energy (e.g. https://www.opencompute.org/circular-economy)
43. • Riekstin, A.C et al. 2018: A Survey on Metrics and Measurement Tools for Sustainable Distributed Cloud Networks, IEEE Communications Surveys & Tutorials • Clicking Clean (Greenpeace), 2017: https://www.greenpeace.org/usa/global-warming/click-clean/ • Google Sustainability Report 2019: https://sustainability.google/reports/environmental-report-2019/ • Facebook PUE/WUE dashboards: https://engineering.fb.com/data-center-engineering/open-sourcing-pue-wue-dashboards/ • D. Dudkowski and K. Samdanis, 2012: Energy consumption monitoring techniques in communication networks, IEEE CCNC • Electricity Map: https://www.electricitymap.org/ • Riekstin, A.C 2015: Orchestration of energy efficiency capabilities for a sustainable network management, Thesis • Georgiou et al. 2019: Software Development Lifecycle for Energy Efficiency: Techniques and Tools, ACM Computing Surveys • Bozzelli et al. 2014: A systematic literature review on green software metrics, VU Technical Report • Energy Profiler: https://developer.android.com/studio/profile/energy-profiler • MLCO2: https://mlco2.github.io/impact/ • Kambadur and Kim, 2014: An experimental survey of energy management across the stack, ACM OOPSLA’14 • Wiki Best Practices: https://wiki.cs.vu.nl/green_software/Best_practices_for_energy_efficient_software • GeSI Smarter 2030: http://smarter2030.gesi.org/downloads/Full_report.pdf • Riekstin A.C et al., 2019: Time Series-Based GHG Emissions Prediction for Smart Homes, TSUSC • Microsoft Smart Building: https://www.microsoft.com/en-us/itshowcase/microsoft-uses-machine-learning-to-develop-smart- energy-solutions • Crown2Map: https://crowd2map.org/ • Venkataramani, S. et al., 2015: Scalable-effort classifiers for energy-efficient machine learning • OpenCompute Project, Circular Economy: https://www.opencompute.org/circular-economy