1

Federal Government Initiatives to Reduce Data Center Energy Use

Sustainable Computing from the Desktop to the Datacenter

Andrew FanaraUnited States Environmental Protection Agency

Climate Protection Partnership DivisionENERGY STAR® Product Specification Development

fanara.andrew@epa.gov

2

Broad Themes We Are Paying Attention to

• Awareness of energy issues and concern for environment and climate change growing among individuals and business globally

• Green economy on our door step -- but lacking a stable foundation• Energy demand associated with our growing digital economy

becoming better understood• Opportunity for tech to improve the energy productivity of many

economic sectors (transportation, manufacturing, products, buildings, lifestyles…)

• US (and growing array of other governments) see efficiency, vs.. supply alone, as a key component national energy strategy

• Datacenter efficiency a prime global opportunity

World Primary Energy DemandBy Major Fuel

Fossil fuels account for 80% of world energy demand and will account for 83% of the increase in overall demand from 2004 -- 2030

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

2004 2015 2030

Other Renew ables

Biomass/Waste

Hydro

Nuclear

Gas

Oil

Coal

From the World Energy Outlook 2006, International Energy Agency

Charting Energy Consumption:

A Pattern of Growing Energy Demand

2004 Energy Consumption = ~ 100 quads2004 Energy Expenditures = ~ $910 billion

0

4000

8000

12000

16000

20000

24000

28000

32000

36000

1940 1950 1960 1970 1980 1990 2000 2010

In d u s t ria l = re dT ra n s p o rt a t io n = p u rp leR e s id e n t ia l = g re e nC o m m e rc ia l = b lu e

U .S . E ne rgy C o nsump tio n1 9 4 9 - 2 0 0 4

A ll F ue ls (T B T U )

Commercial Sector Energy Consumption (quadrillion Btu)

Source: February 2007 Monthly Energy Review, EIA

Electric Power Sector Energy Use (quadrillion Btu)

Source: February 2007 Monthly Energy Review, EIA

Rising Utility Rates2005 Average by Region

8

The Energy Straightjacket

• Deliverability limitations in all markets for all fuels• Oil market restrained by refining capacity• Coal market restrained by rail & mining capacity• Electricity constrained by available fuel

transmission, rising construction cost – high demand taxes grid infrastructure

• Renewables limited by equipment manufacturing• Fuel switching limited by tight markets

9

Energy Efficiency: Still the Cheapest Resource

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.00

0 10 20 30 40 50

Levelized Cost of Electricity by Source

Wind @ 29% CF

IGCC w/o CCS

NGCC @ $6 gas

Biomass

NuclearPulverized Coal w/o CCSEnergy Efficiency

Cen

ts p

er K

Wh

Carbon price: Dollars per ton

Source: Noé van Hulst, IEA

Contributing Factors in CO2 Reductions 2004-2030

Energy-Efficient Investments Could be an Even Bigger Business

• ACEEE estimates that, with additional policies and incentives, the rate of decline in energy intensity could grow from a projected rate of 1.8% to 2.5% per annum through 2030.

• If we were to achieve the higher rate of improvement in intensity, annual investments in energy-efficiency technologies could double to $400+ billion or more annually.

• Increased efficiency would reduce projected energy consumption by ~30% at an average payback of ~5 years (perhaps less).

» Source: ACEEE, American Council for an Energy Efficient Economy

• Increased media attention reflects “resurgence in commitment to environmentalism”

Climate Concerns Gaining Momentum

NY Times, February 6, 2007

CNN, April 29, 2007BBC News, October 31, 2006

14

What are the Risk to Business?

• Energy Supply, Security & Climate Change

Physical risk to property from extreme weather Financial risk to the health and competitiveness of

firms Reputational risk due to poor public and investor

community perception

• There is a growing demand for energy management strategies designed to mitigate that will provide a competitive advantage

15

Where do INFORMATION FACTORIES Fit In?

• Data centers are energy intensive facilities– Server racks now designed to carry 25 kW load– Surging demand for data storage– Typical facility ~ 1MW, can be > 20 MW – Nationally 1.5% of US Electricity consumption in 2006

• Could double in next 5 years

• Critical national and global infrastructure– Few options to go “off the grid” or diversify supply– Good candidates for efficiency investments by utilities to reduce

peak loads

• DC operators want help with their energy challenges!

16

Where do INFORMATION FACTORIES Fit In?

• Significant data center building boom, – Power and cooling constraints in existing

facilities– Growing demand for compute cycles– Growing computing performance – Commoditized hardware– Declining cost of computing

17

Air Movement12%

Electricity Transformer/

UPS10%

Lighting, etc.3%

Cooling25%

IT Equipment50%

Source: EYP Mission Critical Facilities Inc., New York

Other than a common power source they are not connected.

Where Data Center Power Goes

18

Power, Space & Cooling

• Over the next five years, power failures and limits on power availability will halt data center operations at more than 90% of all companies

(AFCOM Data Center Institute’s Five Bold Predictions, 2006)

• By 2008, 50% of current data centers will have insufficient power and cooling capacity to meet the demands of high-density equipment

(Gartner press release, 2006)

• Survey of 100 data center operators: 40% reported running out of space, power, cooling capacity without sufficient notice

(Aperture Research Institute)

19

The Rising Cost of Ownership

• From 2000 – 2006, computing performance increased 25x but energy efficiency only 8x– Amount of power consumed per $1,000 of server

spending has increased 4x

• Cost of electricity and supporting infrastructure now surpasses capital cost of IT equipment

• Perverse incentives -- IT and facilities costs separateSource: The Uptime Institute, 2007

Industry Action: Climate Savers

• Global non-profit consortium of industry, business, universities, conservation groups, governments, consumers

• Goals:– Accelerate production/distribution of energy efficient computers– Increase use of power management tools

• Desired Results:– Reduce the computing industry’s carbon footprint– Lower TCO for computer users– Make high efficiency the norm for the industry

• Web site: www.climatesaverscomputing.org.

Source: Bill Weihl, Google, Digital Power Forum 2007

Massachusetts Institute of Technology Source: Bill Weihl, Google, Digital Power Forum 2007

Industry Action: Green Grid

• Global consortium dedicated to developing and promoting energy efficiency for data centers by:• Defining meaningful, user-centric models and metrics• Developing standards, measurement methods, best

practices and technologies to improve performance against the defined metrics

• Promoting the adoption of energy efficient standards, processes, measurements and technologies

Green Grid Initiatives• Create shared definitions, benchmarks and metrics to

enable real-time measurement monitoring and control of data center efficiency and productivity

• Create baseline 'state-of-the-industry' documentation including benchmark architectures and a repository of data center efficiency knowledge

• Create a comprehensive technology roadmap for future data center design to maximize efficient and productive operations

• Assess new and alternate data center technologies • Monitor progress on all fronts/provide periodic updates

24

What’s the Government’s Role?

• Federal agencies (EPA & DOE) can be catalyst – Stimulate competition on energy efficiency – Foster discussions between key stakeholders– Provide key recommendations (EPA Report to Congress)– Encourage development of standardized test procedures

and metrics to measure energy consumption and make it more transparent

• Promote initiatives globally – Canada, EU, UK, China, India, Australia are showing interest

25

Public Law 109-431: EPA Report

• Purpose: assess energy impacts on and from datacenters, identify energy efficiency opportunities, and recommend strategies to drive the market for efficiency

• Goals:– Inform Congress & other policy makers of important

market trends, forecasts, opportunities– Identify and recommend potential short and long term

efficiency opportunities and match them with the right policies

– Identify areas for additional strategic research outside the scope of the report

26

EPA Report Findings

Trends in Data Center Energy Use • Sector consumed about 61 billion KWh in 2006

– Equates to ~1.5% total U.S. electricity consumption and ~$4.5 billion

– Federal sector: ~6 billion kWh and ~$450 million

• Projected to increase to 100 billion kWh in 2011– Equates to ~2.5% of total U.S. electricity

consumption and ~$7.4 billion

27

Comparison of Projected Electricity UseAll Scenarios 2007 - 2011

28

Electricity Use by End-Use2000 to 2006

29

Report Findings cont.

Identified Key Barriers to Energy Efficiency• Lack of efficiency definitions for equipment and data centers

– Service output difficult to measure, varies among applications– Need for metrics and more data: How do we account for

computing performance?

• Split incentives– Disconnect between IT and facilities managers

• Risk aversion– Fear of change and potential downtime – energy efficiency

perceived as a change with uncertain value and risk

30

Report Recommendations

• Standardized performance measurements for IT equipment and data centers– Development of benchmark/metric for data centers– ENERGY STAR label for servers, considering storage and network

equipment

• Leadership by federal government– Publicly report energy performance of datacenters – Conduct energy efficiency assessments in all datacenters in 2-3

years– Architect of the Capital, implement server-related recommendations

in Greening of the Capital report

31

Recommendations cont.

• Private Sector Challenge– CEOs conduct DOE Save Energy Now energy efficiency

assessments, implement measures, and report performance

• Information on Best Practices– Raise awareness and reduce perceived risk of energy efficiency

improvements in datacenter– Government partner with private industry: case studies, best

practices

• Research and Development– Develop technologies and practices for datacenter energy

efficiency (e.g., hardware, software, power conversion)

32

ENERGY STAR for Servers

• Server energy demand drives DC power & cooling needs• Goal: Create protocol to measure server energy efficiency to

allow fair competition• Technical specification would have several key elements:

– Definitions of product categories eligible for ENERGY STAR

– Test procedure to measure energy efficiency & computing performance

– Energy efficiency performance levels • Tier 1 may include power supply efficiency and other criteria but

would be phased out in the longer term • Tier 2, would be holistic system efficiency metric

33

ENERGY STAR for Servers

Power Supply Efficiency -- A Possible Tier 1• Why higher efficiency for server power supplies?

– Common hardware denominator– Lower HVAC costs >> 1 to 1.5 kWh HVAC savings for every kWh

saved at the plug– More computing space – increase computational density– Reduce CO2 emissions [1kWh ~ 1.6 lbs of CO2]

• Test Procedure developed by Electric Power Research Institute (EPRI)

– Testing and verifying power supply efficiency and reliability performance

– Developing recommendations for 80 Plus program

Source: Brian Fortenbery, EPRI, Digital Power Forum 2007

34

Data on Single Voltage Power Supplies for Servers

Efficiency of Single-Output Server Power Supplies

40%

45%

50%

55%

60%

65%

70%

75%

80%

85%

90%

95%

0% 20% 40% 60% 80% 100%

Percent of Rated DC Power Output (%)

Effici

ency

(%

)

Results of 30 different Single and Multiple Output Server PSUs tested at 230VAC

Source: Brian Fortenbery, EPRI, Digital Power Forum 2007

35

Observations on PS Data

• Data shows a large spread of efficiencies at different load points

• Efficiency drops off rapidly below 20% load, but many Server PS operate below 20% - especially in redundant configurations– 10 % test condition included in test procedure

• Single Voltage server supplies generally have higher efficiency than multi-voltage PC supplies– Operates at higher voltage for increased efficiency– Eliminating less efficient 5V and 3.3V buses improves the overall

efficiency of the power supply for the same power rating

36

ENERGY STAR for Servers

Source: Brian Fortenbery, EPRI, Digital Power Forum 2007

Other Possible Tier 1 Requirements• Power Management - What does this mean for servers?• Standardized labels for consumers to compare server

capability/energy use– Similar to DOE labels on white goods

• Standard Ethernet protocol for querying power consumption of server components

• Adaptive speed Ethernet (similar to Version 4.0 ENERGY STAR Computer Specification)

37

ENERGY STAR for Servers

Server Performance Benchmark – A Possible Tier 2• January 2006 SPEC Power and Performance

Committee began development of benchmark for evaluating energy efficiency of servers

• Working prototype has been developed -- final product by the end of 2007

• Could require reporting of SPEC score in Tier 1 to determine applicability to Tier 2

• More info on progress: www.spec.org/specpower

38

Next Steps for Servers

• Draft framework discussion document distributed for stakeholder review – July 2007

• Current: – review of stakeholder comments on framework discussion

document – gathering relevant information

• October 31: ENERGY STAR stakeholder meeting to discuss Draft 1 specification requirements – Following the Uptime Institute 2007 Charette in Santa Fe, NM

October 28-30 www.uptimeinstitute.org/charette• Goal – Tier 1 specification finalized by early 2008

39

Data Center Energy Plan

Measure data center using developed / available metrics

Implement changes in operations (eliminate comatose servers, virtualize and

consolidate existing servers, enable available power management, implement other best

practices)

Alter procurement policies for new hardware to emphasize efficiency

Remeasure regularly to determine savings

A simple standard to assess data center energy efficiency

• Compares total power used by the data center to the power used by the technology

• Provides a market comparison

• Demonstrates range for opportunity improvement

(1) Standard supported by Green Grid, originally published by Lawrence Berkley National Laboratory, 2003

Most energy efficient

Least energy efficient

1.5 3.52.0 2.5 3.0

CurrentObjective

Data center energy efficiency assessment

41

Opportunities for Energy Savings

• Government will encourage industry to look at all opportunities to improve energy productivity in best practices in DC design, operation and for equipment

• Software will be looked at as both a driver of and a solution to increased energy consumption in the data center

• With more energy transparency due to common metrics, markets can develop to foster fair competition

42

Power Mgnt. for Your Desktop

• Get organizations to…– Primary: Activate power management features (will have

numeric goals)– Secondary: Buy ENERGY STAR products– Think about the impact of IT/office equipment on global warming

• Tie together the various ES IT savings opportunities (products & tools) into something larger and more useful to participants– Offer more ways to save energy & money than just CPM– Leverage corporate carbon reduction initiatives– Coordinate with Green Grid and Climate Savers to carry

message

43

Final Thoughts and Take Aways

• More attention being paid to the implications of energy supply and demand than ever before

• We are seeing the emergence of a more environmentally sensitive consumer class

• Is government leading or following public sentiment?

• Enterprise wide strategic energy management and planning a competitive must have for every organization

• Voluntary and industry schemes (and claims) to save energy will increasingly be questioned for their effectiveness -- they must be challenging and verifiable to be credible

44

Andrew Fanarafanara.andrew@epa.gov

For more information:www.energystar.gov

http://www1.eere.energy.gov/industry/saveenergynow/http://hightech.lbl.gov/datacenters.html

45

46

Energy Efficiency Gives More!

50

100

150

200

1970 1976 1982 1988 1994 2000 2006

En

ergy

Ser

vice

s in

Qu

ads

• Since 1970, energy efficiency has met 77% of new energy service demands in the U.S, while new energy supplies have contributed only 23% of new energy service demands.

Energy Service Demand

Energy Supply

1970 Energy Usage

1

Diagnose

Get the facts to understand your energy use and opportunities for improvement

2 3

Implement

Implement

virtualized

infrastructure

and exploit

innovative

technologies

BuildPlan,

build, and

update

energy

efficient

data

centers

4

Manage &

Measure

Seize control

with power

management

software

5

Cool

Exploit liquid

cooling

solutions –

inside and

outside of

the data

center

1 2 3 4 5

Market is primed for companies capable of integrating technology, services, and industry expertise to help manage and use energy in a more

cost-effective, efficient, and environmentally-sustainable way

A Roadmap to Better Datacenter Mgnt.?

Power use

Chart and data source: American Power Conversion Corporation white paper, Implementing Energy Efficient Data Centers, by Neil Rasmussen, 2006

35

30

25

20

15

10

5

0

Chiller/cooling

tower

Informationtechnology

LightingSwitch/gen

Uninterruptiblepower

supply (UPS)

Power distribution

unit

Computerroom air-

conditioner

Humidifier

Cooling systems Electrical and building systems%

of

tota

l d

ata

ce

nte

r e

lec

tric

ity

us

e