<img alt="" src="https://secure.bomb5mild.com/193737.png" style="display:none;">

Turbonomic Blog

Sustainability calculator: It starts with you. Estimate your cloud and data center carbon footprint today.

Posted by Asena Hertz on Apr 22, 2022 1:22:10 PM
Asena Hertz
Find me on:

Winding Road through forest

Data centers globally use an estimated 200 terawatt hours (TWh) of electricity each year. If the collective data centers of the world were a country, it would rank #23 on a list of highest electricity consumers by nation. Put another way, that’s equivalent to nearly 40% of Germany’s electricity consumption, or 5% of the United States.’ But what’s your IT organization’s impact? What can you do about it?

Today we’re releasing a calculator to help you estimate the carbon emissions associated with your data center and public cloud estates. Our hope is that seeing the numbers—estimates as they are—and their equivalents in terms of cars on the road, what it takes to power households, barrels of oil, etc. will inspire some action.

Home Screen of Turbonomic sustainability calculator

Check out the new sustainability calculator here.

IT’s opportunity: consume data center and cloud resources responsibly

IT has been paving the way to our digital future for decades. Recognizing the environmental impact of the energy it consumes, our industry has made great strides in more efficient hardware, heating & cooling technologies, and investing in and purchasing renewables. Where there is still significant room for improvement is our consumption habits—namely, the all-too-common practice of overprovisioning whether you’re in the data center or the cloud.

Over the years we’ve seen many organizations operating at utilization rates of anywhere from 20-40% in their data centers, while being anywhere from 30-50% overprovisioned in the cloud. While some of this can be attributed to resiliency strategies, you’d be hard-pressed to find an IT organization that will tell you they are satisfied with these numbers. What is increasingly becoming apparent to them is that the costs of not addressing this overconsumption isn’t just financial, but environmental as well.

Business executives see CIOs as critical to achieving sustainability goals. Yet according to IDC*, CIOs rank 6th among executives responsible for the sustainability strategy, implementation, and related purchasing decisions in their organization. We believe this has more to do with limited data, tools, and information available to IT, than any lack of willingness or desire to make a difference.

*Source: Who Are the Key Decision Makers Regarding Sustainability/ESG Strategies and Related Purchasing Decisions? Oct 2021 - IDC Survey Spotlight - Doc # US48320221

Our methodology

We share our methodology below in the spirit of paying it forward and welcoming feedback. The data is limited but growing, the research is continuing to evolve, and we thank those that shared their work in the interwebs so we could learn from it and apply it.

Estimating the Cloud Carbon Footprint

We took much of our inspiration from CloudCarbonFootprint.org. The key difference being our user “inputs” must be information that a data center or cloud administrator can reasonably be expected to know off the top of their head—versus leveraging APIs to get more detailed information. Additionally, in the interest of keeping things simple for users we do not account for storage, but as CloudCarbonFootprint.org notes, there is a power component.

#1 Estimate the energy associated with cloud instances…

Total Watt-Hours per Instance = (Average Watts) * (vCPU hours)

where…

Average Watts = Min Watts + Avg vCPU Utilization * (Max Watts - Min Watts)

AWS:

    • Average Min Watts: 0.74
    • Average Max Matts: 3.5

GCP:

    • Median Min Watts: 0.71
    • Median Max Watts: 4.26

Azure:

    • Average Min Watts: 0.78
    • Average Max Matts: 3.76

Source: https://www.cloudcarbonfootprint.org/docs/methodology/#energy-estimate-watt-hours

and…

vCPU hours = (vCPU/instance) * 8760 (hours per year)

… in other words, vCPU hours per instance per year

Then multiply by the number of instances and convert to kilowatt hours…

Total Energy for All Cloud Instances (kWh) = (Total Watt-Hours per Instance * # instances) / 1000

#2 Estimate the total energy consumption by provider…

Total CSP Footprint Energy (kwh) = (Total Energy for CSP Instances) * (CSP Average PUE)

See CloudCarbonFootprint.org for cloud service providers’ average PUE:

  • AWS Average PUE: 1.135
  • GCP Average PUE: 1.1
  • Azure Average PUE: 1.125
#3 Estimate the emissions by cloud provider and sum total…

CSP Emissions (metric tons CO2) = Total CSP Footprint Energy (kwh) * Cloud Region Grid Emission Factor CO2e (metric ton/kWh)

The carbon emissions related to your cloud workloads depends heavily on where those workloads run. We use the grid emission factors for AWS, Azure, and Google as provided by CloudCarbonFootprint.org here.

Total Cloud Emissions (metric tons CO2) = AWS Emissions (metric tons CO2) + GCP Emissions (metric tons CO2) + Azure Emissions (metric tons CO2)

#4 Estimate the emissions equivalents

To calculate the carbon emission equivalencies we use the formulas provided by United States EPA Greenhouse Gas Equivalencies Calculator here. Please note that there will be slight variation between the results you see in the calculator and what is calculated by the EPA’s site because we are using the numbers they have published, which they have rounded.

Equivalent Passenger Vehicles = Total Cloud Emissions (metric tons CO2) * Passenger Vehicles Factor (metric tons CO2e/vehicles/year)

where…

Passenger Vehicles Factor (vehicles/year) = 8.89 × (10^-3) metric tons CO2/gallon gasoline × 11,556 VMT car/truck average × 1/22.5 miles per gallon car/truck average × 1 CO2, CH4, and N2O/0.993 CO2 = 4.60 metric tons CO2e/vehicle/year

Equivalent Barrels of Oil = Total Cloud Emissions (metric tons CO2) * Barrels of Oil Factor

where…

Barrels of Oil Factor = 5.80 mmbtu/barrel × 20.31 kg C/mmbtu × 44 kg CO2/12 kg C × 1 metric ton/1,000 kg = 0.43 metric tons CO2/barrel

Equivalent Home Electricity Use = Total Cloud Emissions (metric tons CO2) * Home Electricity Use Factor

where…

Home Electricity Use Facto r= 11,880 kWh per home × 947.2 lbs CO2 per megawatt-hour generated × 1/(1-0.073) MWh delivered/MWh generated × 1 MWh/1,000 kWh × 1 metric ton/2,204.6 lb = 5.505 metric tons CO2/home

Equivalent Smartphones Charged = Total Cloud Emissions (metric tons CO2) * Smartphones Charged Factor

where…

Smartphones Charged Factor = 0.012 kWh/charge x 1,562.4 pounds CO2/MWh delivered electricity x 1 MWh/1,000 kWh x 1 metric ton/2,204.6 lbs = 8.22 x 10^(-6) metric tons CO2/smartphone charged

Estimating the Data Center Carbon Footprint

We took much of our inspiration from CloudCarbonFootprint.org. The key difference being our user “inputs” must be information that a data center or cloud administrator can reasonably be expected to know off the top of their head—versus leveraging APIs to get more detailed information. Additionally, in the interest of keeping things simple for users we do not account for storage, but as CloudCarbonFootprint.org notes, there is a power component.

#1 Estimate the energy associated with physical hosts…

Total Energy for All Hosts (Watts) = # hosts * Average Energy / Host (Watts)

where…

Average Energy / Host (Watts) = Server Min Watts + Avg CPU Utilization * (Server Max Watts – Server Min Watts)

In our calculator, the user chooses their server type from this SPEC list, which notes Min/Max Watts: All Published SPECpower_ssj2008 Results

The user also providers their estimated average percent utilization.

Then we convert to kilowatts…

Total Energy for All Hosts (kW) = Total Energy for All Hosts (Watts)/1000

Then we estimate the total energy (kWh) for a year…

Total Data Center Energy (kWh) in a Year = Total Energy for All Hosts (kW) * 8760 hours / year * Average PUE

An average PUE of 1.67 is used in our calculation. This 451 report notes: "Adding it all up, the power usage effectiveness (PUE, the ratio of total datacenter power and the IT power) ranged from 1.63 to 1.70 in the survey sample and was consistent across company sizes. This is in line with broader industry data collected by the Uptime Institute but well below the efficiency of AWS operations." - 451 Research Report. We use the average of 1.63 and 1.70, which is 1.67.

#2 Estimate the emissions associated with the data center footprint…

Total Data Center Emissions (metric tons CO2) = Total Data Center Energy (kWh) / year * Emissions Factor

where the Emissions Factor is…

1,562.4 lbs CO2/MWh × (4.536 × 10^(-4) metric tons/lb) × 0.001 MWh/kWh = 7.09 × 10^(-4) metric tons CO2/kWh

OR

0.000709 (rounded)

Note: Emission equivalents vary slightly vs. EPA calculation because we are using the rounded numbers of the emissions factor (7.09x10^(-4)).

#3 Estimate the emissions equivalents

See #4 (above) in Estimating Cloud Carbon Footprint.

Total Estimated Carbon Footprint & Equivalencies
#1 Sum total data center and cloud emissions…

Total Emissions (metric tons CO2) = Total Data Center Emissions (metric tons CO2) + Total Cloud Emissions (metric tons CO2)

#2 Estimate the carbon sequestration equivalents…

Equivalent Carbon Sequestered by Acres of Forest in 1 year (metric tons CO2 / acre) = Total Emissions (metric tons CO2) * Acres of Forest Sequestration Factor

where…

Acres of Forest Sequestration Factor = -0.23 metric ton C/acre/year* × (44 units CO2/12 units C) = -0.84 metric ton CO2/acre/year sequestered annually by one acre of average U.S. forest

Equivalent Carbon Sequestered by Tree Seedlings Grown for 10 years (metric ton CO2 / tree) = Total Emissions (metric tons CO2) * Tree Sequestration Factor

where…

Tree Sequestration Factor = (-36.4 lbs C/tree × (44 units CO2/12 units C) × 1 metric ton/2,204.6 lbs) = -0.060 metric ton CO2 per urban tree planted

 

For questions, comments, and feedback on our calculator, please contact asena.hertz@ibm.com 


Topics: Sustainability

Try Turbonomic

Looking to gain a better understanding of how Turbonomic works in a sandbox environment?  Check out our self-service demo that you can explore at your own pace.

Try Turbo