Data Center Facility Sustainability Metrics (OCP Data Center Facility white paper, October 2, 2024)

Overview

The paper starts from the observation that 15 years of energy efficiency gains are not enough to meet corporate sustainability goals, and that greenhouse gas emissions, water and waste also need measurement and action plans. It offers a framework for internal tracking and benchmarking across a fleet, and for external use in the OCP Ready Data Center Recognition Program, for regulators such as municipalities and utility commissions, and for tenants evaluating a facility. Metrics are designated Required, meaning measured and published, or Recommended, meaning voluntary, with a suggestion to revisit the designations every year. Each metric also carries a lifecycle phase: design and build, operate, or end of life. The paper states the metrics focus on environmental sustainability and are not comprehensive.

Five categories are defined. Energy has total energy, PUE, total renewable energy, renewable energy factor (REF) and energy reuse factor (ERF), all Required. PUE is defined per ISO/IEC 30134-2 as total energy over IT energy, with three categories according to where IT load is measured (UPS output, PDU or branch circuit, or equipment input); PUE1 is required and the highest category achievable is recommended. The paper cautions that PUE reflects only energy, varies by location and design, and should not be used to compare data centers. REF is renewable energy over total energy, and ERF is energy reused outside the boundary, for example district or greenhouse heating, over total energy.

Greenhouse gas metrics cover Scope 1, Scope 2 location-based and market-based, Scope 3, carbon usage effectiveness (CUE, in kgCO2e per kWh), offsets and hourly renewable matching; Scope 3 and offsets are Recommended. Water metrics cover site withdrawal, site consumption, source energy water, water usage effectiveness (WUE in m3/MWh) and supply chain water. WUE is offered at three levels: WUE1 equals withdrawal, WUE2 subtracts non-industrial reuse, and WUE3 also credits rainwater and industrial reuse and adds water consumed off-site for power. The paper cautions that finding a reliable energy water intensity factor is the hardest part of source water reporting and that WUE comparisons across sites or seasons need care.

Waste metrics cover total waste, e-waste, battery waste and diversion rates, all Recommended. Local ecosystem metrics cover land use, land use intensity in kW/m2, outdoor noise in dB(A) and mean species abundance. Regional notes summarise EU and UK regulation and the draft EU reporting scheme for PUE, WUE, ERF and REF. Future views note that liquid cooling savings inside the IT, such as fans and power supplies, are invisible to PUE, point to the LBNL Total Usage Effectiveness metric, discuss AI data centers, and warn that compressor-less adiabatic cooling may raise water use. Reporting guidance requires annual external reporting, recommends quarterly trailing 12 month reporting, and states facility boundary, period and methodology.

Key points

1. Metrics are designated Required (must be measured and published) or Recommended (voluntary), with annual review of the designations suggested.
2. Required energy metrics are total energy, PUE, total renewable energy, REF and ERF; required GHG metrics include Scope 1, Scope 2 location-based, Scope 2 market-based and CUE.
3. PUE is per ISO/IEC 30134-2 with three categories by measurement point: UPS output (PUE1), PDU or branch circuit (PUE2), equipment input (PUE3); PUE1 is required.
4. The paper says PUE is not intended for comparing different data centers because it varies with location and design.
5. REF is renewable energy divided by total energy; ERF is energy reused outside the data center boundary divided by total energy.
6. WUE is given in m3/MWh and offered at three levels: WUE1 (withdrawal), WUE2 (minus non-industrial reuse) and WUE3 (adds rainwater credit, off-site power water and reuse).
7. Required water metrics are site water withdrawal and WUE; site consumption and source energy water are Recommended.
8. Waste and local ecosystem metrics (waste diversion, land use intensity in kW/m2, outdoor noise in dB(A), mean species abundance) are all Recommended.
9. External reporting is annual (required) and quarterly on a trailing 12 months (recommended); internal tracking should be continuous or monthly.
10. Liquid cooling savings inside IT equipment are not captured by PUE; the paper points to Total Usage Effectiveness as an alternative.

Using it in design

• Build the metering plan so that PUE can be reported at the highest category possible, with IT load measured at PDU or equipment input rather than the UPS output where feasible.
• Plan separate metering for renewable energy, reuse energy export and water withdrawal and consumption, since these feed REF, ERF and WUE.
• State the WUE level (1, 2 or 3) and the energy water intensity factor source when reporting water, and avoid comparing sites on WUE alone.
• Include design-phase metrics such as land use, land use intensity and noise in the project brief for the design and build phase.
• Define the facility boundary, reporting period and methodology up front, especially for mixed-use campuses.
• Where liquid cooling is planned, track total energy and water together, since compressor-less adiabatic cooling can save energy yet raise water use.

Who should read it

Data center owners, designers and sustainability officers who need a consistent KPI set for internal benchmarking and external recognition or regulatory reporting.

Document contents

1. Executive Summary
2. 1. Introduction
3. 2. Compliance with Open Compute Project Tenets
4. 3. Stakeholders and Users
5. 4. Metrics
6. Energy
7. Greenhouse Gases (GHG)
8. Water
9. Waste
10. Local Ecosystem
11. 5. Regional Considerations
12. 6. Future Views
13. Measuring the Impact of Liquid Cooling
14. The Sustainability of AI Data Centers
15. Leveraging AI to Make Data Center Facilities More Sustainable
16. 7. Reporting
17. Reporting Frequency
18. Methodology & Assumptions
19. 8. Conclusion

Read the document

Publication details

Authors

Sean Barlett; Robert Bunger; Jaime Comella; Mark Dansie; Spiros Liolis; Kially Ruiz

Publisher

Open Compute Project Foundation

Year

2024

Document language

English

Source type

Industry body

Access

Open licence — downloadable

Licence

CC BY-SA 4.0 — share with attribution under the same licence

Licence evidence

www.opencompute.org

Citation