Strategies for Improving the Sustainability of Data Centers via Energy Mix, Energy Conservation, and Circular Energy

Overview

ICT is increasingly permeating daily life and more of our data goes to the cloud; events like COVID-19 put an exceptional burden on ICT, which increases implementation and use of data centres and so energy use and environmental impact. The review first presents estimated energy use and emissions, then reviews strategies for energy efficiency and conservation; key elements are innovative cooling techniques, natural resources, automation, low-power electronics and equipment with extended thermal limits. Estimated data-centre energy split: cooling and servers about 43% each, storage drives 11% and network devices 3%; an alternative estimate is roughly 52% ICT, 38% cooling system and 10% other equipment.

On power, for an 800 kW load a 4 × 250 kW internally modular UPS runs at 80% load (small losses), a 5 × 250 kW modular-redundant UPS at 64%, and a 3 × 500 kW parallel-redundant UPS at 53% (high losses); specifying efficiency at 30% load is proposed. IGBT three-level UPS power stages reached about 96% efficiency, SiC transistors exceed 98% almost independently of load, and one SiC-based UPS is reported above 98% for any load above 30%. Only the continuous operating power (COP) generator rating is sufficient for Tier III or IV certification, implying an oversized generator; Tier III is listed at 99.982% availability (1.6 h/yr downtime) and Tier IV at 99.995% (0.4 h/yr).

Reviewed free-cooling studies report PUE of 1.10-1.16 and energy savings of 30%-40%, with some claiming COP up to 9-12 or savings up to 47%-49%; airside economisation needs outdoor air at least 5 °C colder than indoor air. Proper room-layout and ventilation design in large data centres costs an estimated 8%-10% of total cost and usually pays back through energy savings in two to three years; hot-aisle containment is reported as better than cold-aisle containment. White racking instead of black could cut the number of luminaires and lighting energy by as much as 37% each.

A Net Zero Energy Data Center (RenewIT) consumes net zero non-renewable primary energy from grid and district networks over its lifetime; the roadmap runs through efficient envelope, ICT, power distribution and cooling distribution, then reuse of ICT waste heat and renewables. Despite the 'circular' framing it treats circularity as energy reuse and gives no quantitative embodied-carbon or material life-cycle data, which is a limitation.

Key points

1. Estimated data-centre energy split: cooling and servers about 43% each, storage drives 11% and network devices 3%; an alternative estimate is roughly 52% ICT, 38% cooling system and 10% other equipment.
2. For an 800 kW load, a 4 × 250 kW internally modular UPS runs at 80% load, a 5 × 250 kW modular-redundant UPS at 64%, and a 3 × 500 kW parallel-redundant UPS at 53% (high losses); specifying efficiency at 30% load is proposed.
3. IGBT three-level UPS power stages reached about 96% efficiency; SiC transistors exceed 98% almost independently of load; one SiC-based UPS is reported above 98% for any load above 30%.
4. Free-cooling studies report PUE of 1.10-1.16 and energy savings of 30%-40%; airside economisation needs outdoor air at least 5 °C colder than indoor air.
5. Proper room-layout and ventilation design in large data centres costs an estimated 8%-10% of total cost and usually pays back through energy savings in two to three years; hot-aisle containment is reported as better than cold-aisle containment.
6. Only the continuous operating power (COP) generator rating is sufficient for Tier III or IV certification and implies an oversized generator; Tier III 99.982% (1.6 h/yr), Tier IV 99.995% (0.4 h/yr).
7. White racking instead of black could cut the number of luminaires and lighting energy by as much as 37% each.
8. CUE is defined as total CO2 emissions over ICT energy, or equivalently the CO2 emission factor × PUE.

Using it in design

• Calculate UPS load ratio (80% for 800 kW in a modular arrangement, 53% in parallel-redundant) and losses at design stage; consider specifying UPS efficiency at 30% load and compare a SiC-based power stage (above 98%).
• For a Tier III or IV target, set the generator rating on the continuous operating power (COP) basis and add to the cost calculation that this implies an oversized generator.
• Allocate 8%-10% of total cost to room layout and ventilation design and consider hot-aisle containment; payback is two to three years.
• For net-zero-energy targets, reduce load first (envelope, ICT, power and cooling distribution), then reuse waste heat and integrate renewables.

Who should read it

Designers and managers surveying physical-layer energy-saving options (power, cooling, lighting, electronics) for data centres.

Document contents

1. 1. Introduction
2. 2. Power Supply and Continuity (2.1 Tier Classification, 2.2 Power Losses, 2.3 UPS)
3. 3. Energy Conservation of Computer Rooms (3.1 Location, room configuration, room temperature, airflow pattern and active cooling, innovative cooling systems, natural cold source, free cooling, liquid cooling; 3.2 Improvement of Lighting System: lighting control, light sources, other strategies)
4. 4. Electronics and Other Strategies (4.1 Low-Power Servers, 4.2 The Little Box Challenge, 4.3 Direct-Current Power Supply, 4.4 Semiconductors, 4.5 Automation, Monitoring, and Robotization, 4.6 Modular Data Centers)
5. 5. Regulatory Environment Governing Data Centers (5.1 Metrics: indicators, trends; 5.2 Regulations, 5.3 Certifications and Initiatives)
6. 6. Net Zero Energy Data Center
7. 7. Conclusions

Related standards

• ISO/IEC 30134 (PUE/CUE/WUE/ERF)
• EN 50600
• ISO 8528-1:2018
• IEC 62040-3
• ANSI/TIA-942-B
• ASHRAE TC 9.9 thermal guidelines
• ANSI/ASHRAE 90.4
• Uptime Institute Tier classification

Read the document

Publication details

Authors

Matteo Manganelli; Alessandro Soldati; Luigi Martirano; Seeram Ramakrishna

Publisher

Sustainability (MDPI AG)

Year

2021

Volume

13

Issue

11

Pages

6114

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/su13116114

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation