The Water Footprint of Data Centers

Overview

As the internet and ICT diffuse rapidly, data-centre energy efficiency has been researched extensively but water footprint has received little to no attention. The paper conducts a preliminary WF accounting for cooling and energy consumption, using an Apple facility dataset and a comparison of HVAC options for a Phoenix data centre. The global data-centre water footprint (HVAC plus energy) is estimated at 767-147,810 million m³ per year, giving a WUEsource of 1047 to 151,061 m³/TJ; this is much higher than the 787 m³/TJ reported by Sharma et al., because WF accounting includes the water embedded in electricity generation.

Outbound data traffic carries a water footprint of 1-205 litres per gigabyte; the high end is close to the 214 L footprint of 1 kg of tomatoes. The indirect water footprint (electricity) is typically much larger than the direct footprint, so the authors say the first priority for cutting data-centre water footprint should be air-side economisation and energy efficiency.

The result is that the HVAC option with the lowest on-site water use is not always the one with the lowest total footprint; the answer flips with the grid's water intensity. Where WFsource is high, HVAC systems that use some water directly can cut the total footprint (Phoenix case); where WFsource is very low, moving away from direct-water HVAC cuts it. The authors name three key uncertainties for data-centre water-footprint work: the water footprint of electricity, psychrometrics (climate) and the water footprint of the HVAC system.

Limits: the authors stress that their global numbers carry very large uncertainty; values are more indicative of the uncertainties involved than a definite quantification. They call for standard WF/WUE reporting.

Key points

1. Global data-centre water footprint (HVAC plus energy) is estimated at 767-147,810 million m³ per year, giving a WUEsource of 1047 to 151,061 m³/TJ.
2. The WUEsource range is much higher than the 787 m³/TJ reported by Sharma et al., because WF accounting includes the water embedded in electricity generation.
3. Outbound data traffic carries a water footprint of 1-205 litres per gigabyte; the high end is close to the 214 L footprint of 1 kg of tomatoes.
4. The indirect water footprint (electricity) is typically much larger than the direct footprint, so the first priority should be air-side economisation and energy efficiency.
5. Where WFsource is high, HVAC systems that use some water directly can cut the total footprint (Phoenix case); where WFsource is very low, moving away from direct-water HVAC cuts it.
6. Three key uncertainties: the water footprint of electricity, psychrometrics (climate) and the water footprint of the HVAC system.

Using it in design

• Compare HVAC options not only by on-site water use but by total water footprint including the grid's water intensity (WUEsource); the ranking can flip with the grid.
• Consider air-side economisation and energy efficiency as the first step to cut water footprint, because electricity-related indirect water usually dominates.
• Report water-footprint results as ranges with the uncertainties of electricity water intensity, climate and HVAC water footprint.

Who should read it

Designers and sustainability managers doing data-centre siting, cooling selection and water-footprint reporting.

Document contents

1. 1. Introduction
2. 2. Water Footprint for Measuring Water Use Efficiency
3. 3. Data Center Structure and Water Footprint of Its Components (3.1 IT Equipment and Uninterruptible Power Supply, 3.2 Direct WF—HVAC, 3.3 Psychrometrics: Temperature and Humidity, 3.4 Indirect WF—Energy Source)
4. 4. Uncertainties (4.1 Uncertain WFs of HVAC Systems, 4.2 Water Footprint of Data Centres and Data Centre Outbound Traffic)
5. 5. Caveats for Policy and Innovation Recommendations
6. 6. Conclusions

Related standards

• ISO/IEC 30134-9 (WUE)
• The Green Grid WP35 (WUE)

Read the document

Publication details

Authors

Bora Ristic; Kaveh Madani; Zen Makuch

Publisher

Sustainability (MDPI AG)

Year

2015

Volume

7

Issue

8

Pages

11260-11284

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/su70811260

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation