Report to Congress on Server and Data Center Energy Efficiency, Public Law 109-431 (August 2, 2007)

Overview

The report is a study that EPA prepared at Congress's request with a study team led by Lawrence Berkeley National Laboratory. The team held a public workshop on February 16, 2007 with about 130 attendees and took comments on the public draft from more than 50 organisations. It finds that U.S. servers and data centers used about 61 billion kWh in 2006, roughly 1.5% of national electricity use and about $4.5 billion in cost. That is more than double the 2000 level.

The method is bottom-up: for each year it estimates the installed base of servers, external disk drives and network ports in data centers, multiplies by unit annual energy use, and scales the sum with an overhead factor for power and cooling. That factor is a PUE of 2.0, based on LBNL benchmarking of 22 data centers. In 2006 infrastructure accounted for 50% of total energy, volume servers for 68% of IT equipment electricity, and enterprise-class data centers for 38% of total consumption.

If current efficiency trends continued, consumption was expected to exceed 100 billion kWh by 2011, annual cost to reach $7.4 billion and peak grid load to rise from about 7 GW to 12 GW. Federal servers and data centers used about 6 billion kWh (10%) and spent about $450 million per year. The report stresses that these figures rest on limited data and that developments such as server consolidation add uncertainty.

Three efficiency scenarios are compared. Improved operation saves more than 20% against current trends in 2011, best practice up to 45% and state of the art up to 55%, equal to annual savings of 23 to 74 billion kWh and 15 to 47 million metric tons of CO2. Infrastructure assumptions: 30% less infrastructure energy from airflow management, up to 70% for best practice, up to 80% for state of the art. Capital measures were applied to only 50% of existing data centers, and IT equipment was assumed to turn over within five years.

The report also examines barriers (no efficiency definition, split incentives, risk aversion), distributed generation and combined heat and power, existing programmes and policy recommendations. First among the recommendations are an objective energy-performance rating system for data centers and server performance metrics. The figures date from 2007 and actual growth diverged from these projections, so the document is best read for historical context and scenario assumptions, not as a current benchmark.

Key points

1. U.S. servers and data centers consumed about 61 billion kWh in 2006, about 1.5% of national electricity and about $4.5 billion.
2. Power and cooling infrastructure was about 50% of total data center energy in 2006; the baseline assumption is an average PUE of 2.0 (from benchmarking of 22 data centers).
3. Federal servers and data centers used about 6 billion kWh (10% of the national total), about $450 million per year.
4. Under current trends, consumption was projected to exceed 100 billion kWh by 2011 at a cost of $7.4 billion a year.
5. Data center peak grid load was about 7 GW (about 15 baseload power plants), expected to reach 12 GW by 2011.
6. Volume servers used 68% of IT equipment electricity in 2006, and enterprise-class data centers 38% of total consumption.
7. Scenario savings in 2011 against current trends: improved operation more than 20%, best practice up to 45%, state of the art up to 55%.
8. The best-practice scenario assumes PUE of 1.5 by 2011 with 98% efficient transformers and 90% efficient UPS; state of the art assumes PUE 1.4 for enterprise sites with 95% efficient UPS and liquid cooling to the racks.
9. The improved-operation scenario assumes PUE of 1.7 by 2011 with 95% efficient transformers and 80% efficient UPS.
10. The report recommends an objective energy-performance rating for data centers and separate utility meters; its data are from 2007 and actual growth diverged from the projections.

Using it in design

• Use it for historical context on PUE: knowing that 2.0 was the 2007 average starting point, compare your current target PUE values against that reference.
• Read the infrastructure improvement steps as a checklist: airflow management, transformer and UPS efficiency, efficient chillers, fans and pumps, free cooling, direct liquid cooling and combined heat and power.
• Compare the transformer and UPS efficiency assumptions (for example 98% and 90% or 95%) with the loss assumptions in your own design calculations; the report presents them as historical scenario parameters.
• Include separate metering of electricity at large sites and separate monitoring of IT and infrastructure energy in the design; the report treats it as basic to both estimation quality and policy.
• When weighing distributed generation or CHP, recall the report's caution: technologies such as fuel cells had no long track record in high-availability applications, and proven availability data are needed.

Who should read it

Designers, policy and strategy teams who want the historical context of data center energy use, the origins of PUE and the assumptions behind efficiency scenarios.

Document contents

1. Executive Summary
2. Introduction
3. Trends in Growth and Energy Use Associated with Servers and Data Centers in the U.S.
4. Estimates of U.S. Server and Data Center Energy Use
5. Energy Use Associated with Federal Government Servers and Data Centers
6. Potential Energy and Cost Savings through Improved Energy Efficiency
7. Electric Utility Impacts from Energy Efficiency in Servers and Data Centers
8. Potential Impacts of Energy Efficiency on Product Performance, Reliability, Features, and Overall Cost
9. Distributed Generation and Combined Heat and Power Systems in Data Centers
10. Current Energy Efficiency Programs Applicable to Data Centers
11. Recommendations for Incentives and Voluntary Programs
12. Recommendations for Research and Development Activities

Read the document

Publication details

Authors

U.S. EPA ENERGY STAR Program

Publisher

U.S. EPA / ENERGY STAR

Year

2007

Document language

English

Source type

Government / public

Access

Open licence — downloadable

Licence

U.S. public domain — U.S. Government work; free to reuse

Licence evidence

www.copyright.gov

Citation