GreenEx:

Put the Planet into your TCO

​

Bob Flynn, Internet2

Alan Walsh, Indiana University (in absentia)

internet2.edu

Introductions

Bob Flynn

Program Manager

Cloud Infrastructure and Platform Services

Internet2

bflynn@internet2.edu

Alan Walsh

Data Engineer

Research Technologies

Indiana University

alwalsh@iu.edu

[ 2 ]

Title, Header

[ 3 ]

What about sustainability?

[ 4 ]

What is the impact of our personal computing choices?

Do sustainability concerns impact what you buy or how often you upgrade?

Is cost your only driver? Are needs?

[ 5 ]

What is the impact of our institutional computing choices?

Do sustainability concerns impact what you buy or how often you upgrade?

Is cost your only driver? Are needs?

​

What if you had an easy way to factor in sustainability?

[ 6 ]

We have the answer!

https://cloud.withgoogle.com/region-picker/

[ 7 ]

Thank you!

[ 8 ]

What we are not doing

[ 9 ]

What we are doing

• Asking questions to provoke others to start asking questions
• Talk about the known knowns
• Talk about the known unknowns
• Try to determine if the unknown unknowns are knowable

…and important?

[ 10 ]

What we are trying to accomplish?

• Start a conversation
• Get people thinking
• Move us closer to a point where we have true TCO for all compute
• What is the right model for TCO?

​

• Is anyone familiar with Der Grüne Punkt?

[ 11 ]

The “cost” of coffee

$

[ 12 ]

The TCO of coffee

​

[ 13 ]

Breaking down T. C. O.

• What does TCO stand for?
• “O” really means operation since cloud computing involves no ownership.
• “T” is not really meaningful unless you account for…everything
• “C” requires us to think more broadly (social cost)

[ 14 ]

[ 15 ]

Thank you!

[ 16 ]

tl;dr

There is no such thing as “free” computing

[ 17 ]

Challenges

[ 18 ]

Challenge 1: The consequences of incomplete data

​

​

• How many people know the cost of power for their data center?
• And if you don’t know. You don’t have the incentive to improve.

[ 19 ]

Challenge 2: Even if you know the cost of power, what is its environmental impact?

✔

[ 20 ]

Challenge 3: What are the other costs?

​

​

[ 21 ]

Challenge 3: What are the other costs?

​

​

Hardware Lifecycle

[ 22 ]

The Big Takeaway

• You may think your cost is C, but it’s really C+E, with E being the environmental or “social” cost.
• E is hard to pin down, but we can make a start.

[ 23 ]

Equation to Determine E for a Datacenter

[ 24 ]

Simplistic Version of the equation – Focused on 1 GPU

E =

(Total Energy Consumption for a GPU running per hour * Carbon Intensity)

​

(Social Cost Carbon)

In kWh

In $/Carbon Tons

In Carbon Tons/kWh

SCC Range: (110, 190, 330)

*

E =

(.250 * 1.08 * (159.278)) * $190

In kWh

In $/Carbon Ton

=

$0.00817

Total Energy consumption

PUE

Social Cost of Carbon

**Per hour**

Carbon Intensity in

gCO2eq/KWh per state

Need for a simplified equation

While a comprehensive equation to calculate cost to environment is helpful, many situations demand for a much simpler equation that can give an approximation of environmental impact. As a GPU is a fundamental block of a datacenter, its environmental impact can then be used to gauge the impact of a datacenter.

Mockup for GPU (Nvidia A100 40GB) in Texas

1,000,000

~32,000 GPUs limit per data center – Datacenter Dynamics

Mockup for Annual Social Cost for Meta Datacenter in Texas

$60,074,411.84

​

[ 28 ]

E =

(959,419,000) * 1.08 * (241g/1,000,000)) * $190

In kWh

=

$47,446,340

Total Energy consumption

Carbon Emissions (MT)

PUE

Social Cost of Carbon (EPA 2.0)

Mockup for Annual Social Cost for Meta Datacenter in Texas (removing social cost of water)

E =

(21,000,000) * 1.2 * (373g/1,000,000)) * $190

In kWh

=

$1,785,924

Total Energy consumption

Carbon Emissions (MT)

PUE

Social Cost of Carbon (EPA 2.0)

R1 in the Midwest (~20K/sq ft on-prem data center)

For the equivalent power consumption as the Meta data center (959,419,000 kWh) E = $81,592,829 (vs $47,446,340 for Meta)

What do you do with this information?

[ 31 ]

What do you do with this information?

• We need to get computing choice included in institutional sustainability and budgetary thinking
• Find your own TCO

[ 32 ]

Equation combined with TCO

TCO Framework

According to ECAR Working Group Paper, TCO can be broken up into three different factors:

• Quantitative factors,
• Qualitative factors
• Hidden costs/Subsidies

Our team has integrated the sustainability equation into the TCO Framework spreadsheet.

Integrated Spreadsheet

Demonstration of Integrated Spreadsheet

What do you do with this information?

• We need to get computing choice included in institutional sustainability and budgetary thinking
• Find your own TCO

• Consult your office of sustainability
• Start a conversation with the people paying the bills
• Work together to develop incentive structures for greener computing choices

[ 34 ]

Thank you!

(for real this time)

[ 35 ]

Acknowledgements

• IU Kelley School of Business MSIS Capstone Team
• Stephen Kim, Bhargava Mangalagiri, Jake Schoenegge, Adam Warner
• Kelsey Beal and Jessica Davis, IU Office of Sustainability
• IU O’Neil School of Public and Environmental Affairs E555 Lifecycle Assessment (LCA) course
• TechEx 2023: Cloudy with a Chance of Conservation
• Keith Wessel, Rob Carter, Netta Caligari, Scott Woods

[ 36 ]

Acknowledgements

• AWS
• Tommy Johnston, Rick Friedman, Scott Friedman, and John Dittmer
• Google
• Tazzie Green, and Trinity Lloyd
• Microsoft
• Sundar Ramakrishnan
• Crew Universal
• Gim Crew

[ 37 ]

Appendix

[ 38 ]

For Further Research

• General
• TCO for Cloud Services: A Framework | EDUCAUSE Library
• Cloudy With a Chance for Conservation (TX23 Presentation)
• Recalibrating global data center energy-use estimates | Science
• Measuring the Carbon Intensity of AI in Cloud Instances | Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency
• Sustainable Computing Research (several reports and research papers we collected along the way)
• In the News
• AI’s craving for data is matched only by a runaway thirst for water and energy | John Naughton | The Guardian
• Artificial intelligence technology behind ChatGPT was built in Iowa — with a lot of water | AP News
• AI’s craving for data is matched only by a runaway thirst for water and energy
• The Growing Environmental Footprint Of Generative AI
• Markey, Heinrich, Eshoo, Beyer Introduce Legislation to Investigate, Measure Environmental Impacts of Artificial Intelligence
• GreenOps
• Future IT: FinOps, GreenOps and sustainable cloud strategies | Capgemini
• Cloud Sustainability – A Union of FinOps and GreenOps - Xebia
• FinOps and GreenOps Strategies in 2023 - IDC Europe Blog
• FinOps and GreenOps: How do They Relate?
• IT Sustainability Think Tank: Embedding GreenOps into enterprises | Computer Weekly
• What are Scopes 1, 2 and 3 of Carbon Emissions?
• What Is GreenOps? Putting a Sustainable Focus on FinOps - The New Stack

[ 39 ]

For Further Research

• AWS
• 2022 Sustainability Report Highlights
• Cloud powers faster, greener, and more collaborative research, according to new IDC report | AWS Public Sector Blog
• AWS Makes Water Positive Commitment to Return More Water to Communities Than It Uses by 2030
• Reducing carbon by moving to AWS
• Renewable Energy Methodology
• The Cloud - Amazon Sustainability
• Google
• Google Data Centers
• Google Data Center Water Use in the US Revealed To Be a Lot.
• Secret Cost of Google's Data Centers: Billions of Gallons of Water | TIME
• Restoring Ecosystems through Water Stewardship - Google Sustainability
• Achieving Our 100% Renewable Energy Purchasing Goal and Going Beyond
• Cloud sustainability
• Efficiency – Data Centers – Google
• Microsoft
• Microsoft commits to achieve ‘zero waste’ goals by 2030 - The Official Microsoft Blog
• Videos to configure Sustainability Manager (to get the emission report of on-prem and in Azure):
• Get Started with Microsoft Sustainability Manager
• Microsoft Cloud for Sustainability Demo
• Documentation:
• Microsoft Sustainability Calculator helps enterprises analyze the carbon emissions of their IT infrastructure | Azure Blog | Microsoft Azure
• Microsoft Sustainability Manager overview - Microsoft Cloud for Sustainability | Microsoft Learn

[ 40 ]

EPA Social Cost of Carbon

How did they calculate the carbon tax? What was the logic, and how can we apply it to water

​

• There are two ways of taxation:
• Levy on energy suppliers on carbon emissions
• Credits to induce a reduction of emissions
• IMF reports such taxations to help reduce the amount of emissions.
• (The organisation tracks the reductions by years and the price level)

​

• Formula: 
• Amount of Carbon emission (ton) * price
• The expected monetary amount of the taxation: $15/ton �(The Brookings Institute)
• ‘Carbon Pricing’ can be an alternative to taxation rates.

• Carbon taxation implementation in America: CA, MA, OR, PA, WA,
• But it is implemented globally.

The Ideal Equation

Note: WUE is the water used for all the energy used by the equipment in the datacenter. So, I am not sure how it can be incorporated in the first part of the equation

E =

(Total Water Consumption * Water Usage Effectiveness) * Social Cost of Water

(Total Energy Consumption * Power Usage Effectiveness * Carbon Intensity)* Social Cost Carbon

+

In Gallons

In kWh

In Gallons/kWH

In $/Gallon

In $/Carbon Tons

In Carbon Tons/kWh

E =

(Total Water Consumption * Social Cost of Water)

(Total Energy Consumption * Power Usage Effectiveness * Carbon Intensity)* Social Cost Carbon

+

In Gallons

In kWh

In $/Gallon

In $/Carbon Tons

In Carbon Tons/kWh

Inclusion of WUE & its challenges

The Broken-down Equation without WUE

Total Facility Power / IT Equipment Power

E =

Total Water Consumption * Social Cost of Water

(Total Energy Consumption * Power Usage Effectiveness * Carbon Intensity) * Social Cost Carbon

+

In Gallons

In kWh

In $/Gallon

In $/Carbon Tons

In kWh

In kWh

In Carbon Tons/kWh

Water Risk Factor * price per gallon in region

The Broken-down Equation with WUE

Total Facility Power / IT Equipment Power

Water Consumption/ IT Equipment Power

E =

(Total Water Consumption * Water Usage Effectiveness) * Social Cost of Water

(Total Energy Consumption * Power Usage Effectiveness * Carbon Intensity) * Social Cost Carbon

+

In Gallons

In kWh

In Gallons/kWH

In $/Gallon

In $/Carbon Tons

In kWh

In kWh

In Carbon Tons/kWh

Water Risk Factor * price per gallon in region

Explanation of Water variables

• Total Water Consumption (Gallons)
• Facility's water intake per year 
• WUE
• WUE = Data Center Water Consumption / IT Equipment Energy
• WUE is the water used per kilowatt of energy used by the equipment
• Unit is Gallon/kWh
• Sources - https://www.sunbirddcim.com
• Social Cost of water ($ per gallon)
• Price per gallon in region for a data center
• Sources - https://riskfilter.org/water/explore/data-and-methods

Explanation of Energy variables

• Total Energy Consumption (kWH)
• Facility's Energy intake per year
• PUE
• Total facility Power / IT Equipment Power
• Carbon Emissions (grams of carbon per KWH)
• Emissions of Carbon in that area
• Sources - https://app.electricitymaps.com/map?lang=en
• Social Cost of Carbon ($ per carbon ton)
• Price of Carbon Emissions
• Sources - https://www.rff.org/topics/scc/

Sources MSIS team used

2 gallons per KWH (2016)

• https://www.osti.gov/servlets/purl/1372902/

​

Question to be asked:

• Would this only be used if Water Consumption isn't provided?

​

Water Data

Source: WWF Methodology

Source: EPA

Emissions Data

Factors that could impact social cost:

• State water risk 
• State emissions per capita 
• Energy efficiency of data center

PUE is from META datacenter average

Incentives for Universities

• Phase 1: Key Question: how to segregate energy bill based on multiple buildings?
• Approach:
• Build an inventory and classify buildings according to use (academic, residential, data center, laboratories)
• Identify High energy facilities (ie, those that are expected to have higher energy consumption)
• Analyse occupancy and utilization levels to know when to expect peaks and lows in utilization
• Establish baseline expectations based on above data & comparison with similar infrastructure
• Implement sub-metering plan to be able to gauge electricity usage per building
• Establish analytics dashboard to compare consumption data and identify outliers.

Incentives for Universities

• Phase 2: Key Question: how to decrease consumption in buildings with high energy usage?
• Approach:
• Surprises:
• It is expected that labs with energy heavy equipment might use lots of electricity. But if the analysis reveals that administrative buildings are showing comparable electricity usage then this is an area of improvement.
• The cumulative effect of small appliances and personal equipment (like mini-fridges in dorm rooms, microwaves, and personal heaters or fans) can lead to a surprisingly high energy draw. This scenario might highlight the need for policies to manage personal energy uses campus-wide.
• Further investigate those buildings/departments with an IT team
• Qualify what accounts as IT use:
• Just all IT systems that are in the classrooms, offices.
• Or more specifically, the systems used to host IU software (on prem & cloud)
• If this is the case, then seek bills for cloud usage & electricity bill of IU data center
• This part is more easier to investigate

​

Incentives for Universities

• Phase 3: Key Question: how to decrease consumption/ promote sustainable cloud usage in departments high energy usage?
• Approach:
• This is where TCO formula comes into picture
• Can cite studies, surveys:
• https://www.timeshighereducation.com/news/sustainability-more-important-location-mobile-students
• Applicants are just as likely to rate university sustainability and graduate employment prospects as top priorities
• https://sos.earth/survey-2020/
• 92% agree that sustainable development is something which all universities and colleges should actively incorporate and promote
• https://blog.thepienews.com/2023/03/the-importance-of-sustainability-in-students-university-choice/
• The rankings are increasingly taking into account how the universities are working towards the UN SDG
• As the outcome of these surveys and research shows how it can impact the main consumers of universities, ie, the students, it is a big incentive to work towards becoming carbon neutral.

​