A MATURITY MODEL FOR ENERGY EFFICIENCY

IN MATURE DATA CENTRES

Edward Curry

, Gerard Conway

, Brian Donnellan

, Charlie Sheridan

and Keith Ellis

Digital Enterprise Research Institute, National University of Ireland, Galway, Ireland

Innovation Value Institute, National University of Ireland, Maynooth, Ireland

Intel Labs Europe, Intel Corporation, Leixlip, Ireland

Keywords: Data Centre, Energy Efficiency, Maturity Model, Data Centre Management, IT Management.

Abstract: Data centres are complex eco-systems that interconnect elements of the ICT, electrical, and mechanical

fields of engineering and hence the efficient operation of a data centre requires a diverse range of

knowledge and skills from each of these fields. The Innovation Value Institute (IVI), a consortium of

leading organizations from industry, the not for profit sector, and academia, have developed a maturity

model that offers a comprehensive, value-based method for organizing, evaluating, planning, and improving

the energy efficiency of mature data centres. The development process for the maturity model is discussed,

detailing the role of design science in its definition.

1 INTRODUCTION

According to McKinsey & Co. (Forrest and Kaplan,

2008) the world’s 44 million servers consume 0.5%

of all electricity and produce 0.2%, or 80 megatons,

of carbon dioxide emissions a year. Given a business

as usual scenario, by 2020 greenhouse gas emissions

from Data Centres (DCs) are projected to more than

double from 2007 levels (Webb, 2008). The efficient

operation of a data centre requires a diverse range of

knowledge and skills from a large ecosystem of

stakeholders. A DC requires expertise from

engineering (including electrical, civil, mechanical,

software, and electronic) to accountancy to systems

management. The Innovation Value Institute (IVI), a

consortium of leading organizations from industry

(including, Microsoft, Intel, SAP, Chevron, Cisco,

The Boston Consultancy Group, Ernst & Young, and

Fujitsu), the not for profit sector, and academia, has

developed and tested a maturity model for

systematically assessing and improving energy

efficient capabilities within mature DCs. The model

offers a comprehensive, value-based model for

organizing, evaluating, planning, and managing DC

capabilities for energy efficiency and fits within

IVI’s IT-Capability Maturity Framework (IT-CMF)

for managing IT. The model provides a high-level

assessment of maturity for IT managers with

responsibility for DC operations.

2 DC ENERGY COMSUMPTION

Power usage within a DC goes beyond the direct

power needs of servers to include networking,

cooling, lighting, and facilitie. Power draws for DCs

range from a few kilowatts for a rack of servers to

several tens of megawatts for large facilities. While

the exact breakdown of power usage will vary

between individual DCs, Figure 1 illustrates the

examination of one DC where up to 88.8% of the

power consumed by the DC was not used on

computation (U.S. EPA, 2007). International Data

Corporation (IDC) estimates that DC energy costs

will be higher than equipment costs by 2015

(Martinez and Bahloul, 2008). The cost of operating a

DC goes beyond the economic bottom line; there is

also an environmental cost. DCs are the fastest

growing contributor to the IT sectors environmental

footprint and are predicted to grow to 259 MtCO2e by

2020, up from 76 MtCO2e in 2002 (Webb, 2008).

3 DC ENERGY EFFICIENCY

With electricity costs being the dominant operating

expense of a DC, it is vital to maximize the

operational efficiency in order to reduce both the

Environmental and economic cost. Energy efficient

263

Curry E., Conway G., Donnellan B., Sheridan C. and Ellis K..

A MATURITY MODEL FOR ENERGY EFFICIENCY IN MATURE DATA CENTRES.

DOI: 10.5220/0003953702630267

In Proceedings of the 1st International Conference on Smart Grids and Green IT Systems (SMARTGREENS-2012), pages 263-267

ISBN: 978-989-8565-09-9

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Example breakdown of power usage within a Data Centre.

DC operations require a holistic approach to both IT

and facilities energy management. DC and IT

leaders are often unable to find satisfactory answers

to questions such as:

• What is the utilization of the DC?

• How energy efficient is the DC?

• Are there clear measurable goals and objectives

for DC energy efficiency (EE)?

• What is the roadmap for DC EE improvements?

IT departments face additional challenges – such as the

introduction of new methods and tools, and

conformance to industry metrics and standards – which

are compounded by a general lack of relevant

information, such as power consumption

quantifications. Mature DCs typically have a

heterogeneous IT infrastructure with fixed support

systems, making it arduous to employee catch-all

solutions.

4 THE NEED FOR A MATURITY

MODEL

Maturity models are tools that have been used to

improve many capabilities within organizations,

from Business Process Management (BPM)

(Rosemann and De Bruin, 2005) to Software

Engineering (CMMI) (Paulk et al., 1993). Maturity

models have also been developed to support the

management of IT organizations. IVI have

developed the IT-Capability Maturity Framework

(IT-CMF) (Curley, 2004) that provides a high-level

process capability maturity framework for managing

the IT function within an organization to deliver

greater value from IT by assessing and improving a

broad range of management practices. A core

function of IT-CMF is to act as an assessment tool

and a management system.

There is a need to improve the behaviours,

practices, and processes within DCs in order to

deliver greater energy efficiency. To address this

need, the IVI consortium has extended the IT-CMF

with a maturity model for systematically assessing

and improving DC capabilities for energy efficiency.

4.1 Design Methodology

The development of the model was undertaken using

a design process with defined review stages and

development activities based on the Design Science

Research (DSR) guidelines advocated by Hevner et

al. (Hevner, March, Park, and Ram, 2004). During

the design process, researchers participate together

with practitioners within a working group to

research and develop the model. The working group

interviewed multiple DC stakeholders to capture the

views of key domain experts and to understand

current practice and barriers to improving DC

energy efficiency (EE). The working group widely

consulted the relevant literature, both industrial and

academic, on DC EE. Industrial best practices –

including the EU code of conduct for DC EE and the

work of the Green Grid on metrics – were

incorporated. The initial maturity model was

developed in mid-2010 and has been piloted within a

number of DCs, with learning and feedback

incorporated into subsequent versions.

5 MATURITY MODEL

The Data Centre EE model offers a comprehensive,

value-based model for organizing, evaluating,

SMARTGREENS2012-1stInternationalConferenceonSmartGridsandGreenITSystems

264

planning, and managing DC EE capabilities. The

model fits within IT-CMF (Curley, 2004) and is

aligned with the broader Sustainable ICT critical

capability (Curry and Donnellan, 2012; Donnellan,

et al., 2011).

The DC EE assessment methodology determines

how different DC capabilities are contributing to

energy efficiency goals and objectives. The gap

analysis between what energy efficiency targets are,

and what they are actually achieving, positions the

Data Centre EE model as a management tool for

aligning relevant capabilities with EE objectives.

The model focuses on the execution of four key

actions to improve the management of EE in the DC:

• Define goals and objectives for the DC program

• Understand the current DC maturity level,

• Systematically develop and manage the DC

capability building blocks.

• Assess and manage DC progress over time.

5.1 Capability Building Blocks

The Data Centre EE model consists of seven

capability building blocks (see Table 2) in the

categories of Management, Operations, and

Building. The maturity level for each of the seven

Capability Building Blocks is presented in Table 1.

Table 2: Capability Building Blocks of Energy Efficient

Data Centres.

 Capability Description

Management

Organizational

Structure

Howthedatacentreanditsenergyefficiencyis

managed,whoisresponsibleforrunningthe

DC,andhowintegratedare:IT,Facilities,and

theBusiness.

Policy

Thepoliciesinplaceforenergyefficiency

withinthetheDCandhowtheyarealigned

acrosstheenterprise.

Manageability

andMetering

ThemeteringusebyITandFacilitiesto

improveunderstandingandmanageabilityof

energyusage.

perations

IT

Infrastructure

andServices

ThemanagementofITequipmentandservices

toensureenergyefficiency.

Building

InternalAir

andCooling

Theinternalairmanageemnttechniques

employed.

CoolingPlant

Thedesignandmanagementofthecooling

system.

Power

Infrastructure

Themanagementofpowergeneration,and

conditioninganddeliverysystemstomaximize

energyefficiency.

5.2 Assessment Approach

The assessment begins with an online survey of DC

stakeholders in order to understand their individual

assessments of the maturity and the importance of

these capabilities. Typically, a range of individuals

who are involved in, or accountable for, EE for the

DC complete the survey. A series of targeted

interviews with key stakeholders augments the

survey to understand key business priorities and

energy efficiency drivers, successes achieved, and

initiatives taken or planned. Interviews last between

60 and 90 minutes; they are used to support the

survey data.

Figure 2: Reported vs. Desired Maturity.

When the assessment is complete, organizations will

have a clear view of current capability and key areas

for action and improvement. A smaple pilot

assessment result in illustrated in Figure 2. However,

to further develop the capability, the organization

should assess and manage progress over time by

using the assessment results to 1) develop a roadmap

and action plan and 2) add a yearly/half-yearly

follow-up assessment to the overall DC energy

efficiency management process.

6 SUMMARY AND FUTURE

WORK

The IVI consortium has developed and tested a

maturity model for systematically assessing and

improving Energy Efficient capabilities within

mature Data Centres. The resulting model offers a

comprehensive, value-based model for organizing,

evaluating, planning, and improving the energy

efficiency of mature data centres. Further

development and evaluation of the model is planned

– in particular, the use of the model in conjunction

with metrics such as PUE, CUE, WUE, and

CapEx/OpEx costs – in order to quantify benefits.

We are also employing the model to benchmark DCs

AMATURITYMODELFORENERGYEFFICIENCYINMATUREDATACENTRES

265

within the IVI consortium to faciliate comparisons

between DCs across and between industrial sectors.

ACKNOWLEDGEMENTS

We would like to recognize the contribution of the

members of the IVI’s Data Center Energy Efficiency

Working Group. Enterprise Ireland funded part of

the work presented in this paper under Grant

CC/2009/0801.

REFERENCES

Curley, M. (2004). Managing Information Technology for

Business Value: Practical Strategies for IT and

Business Managers (p. 350). Intel Press,.

Curry, E., & Donnellan, B. (2012). Understanding the

Maturity of Sustainable ICT. In J. vom Brocke & S.

Seidel (Eds.), Green Business Process Management –

Towards the Sustainable Enterprise. Springer.

Donnellan, B., Sheridan, C., & Curry, E. (2011). A

Capability Maturity Framework for Sustainable

Information and Communication Technology. IEEE IT

Professional, 13(1), 33-40. doi:10.1109/MITP.2011.2

Forrest, W., & Kaplan, J. M. (2008). Data centers : How to

cut carbon emissions and costs. McKinsey on Business

Technology, 14(6).

Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004).

Design Science in Information Systems Research. MIS

Quarterly, 28(1), 75-105. MIS RESEARCH

CENTER-SCHOOL OF MANAGEMENT.

doi:10.2307/249422

Martinez, N., & Bahloul, K. (2008). Green IT Barometer

European Organisations and the Business Imperatives

of Deploying a Green and Sustainable IT Strategy.

International Data Corporation. Retrieved from

http://content.dell.com/us/en/corp/d/corporate~researc

h~en/Doc uments~idcwp28q.pdf.a

Paulk, M. C., Curtis, B., Chrissis, M. B., & Weber, C. V.

(1993). The Capability Maturity Model for Software.

Software engineering project management,

10(CMU/SEI-93-TR-24 ESC-TR-93-177), 1-26.

Wiley India Pvt. Ltd.

Rosemann, M., & De Bruin, T. (2005). Towards a Business

Process Management Maturity Model. In D. Bartmann,

F. Rajola, J. Kallinikos, D. Avison, R. Winter, P. Ein-

Dor, J. Becker, et al. (Eds.), 13th European Conference

on Information Systems (Vol. 53, pp. 521-532).

Citeseer. doi:10.1007/s11225-008-9119-4

U.S. EPA. (2007). Report to Congress on Server and Data

Center Energy Efficiency Public Law 109-431 (Vol.

109, p. 431).

Webb, M. (2008). SMART 2020: Enabling the low carbon

economy in the information age. The Climate Group

London.

SMARTGREENS2012-1stInternationalConferenceonSmartGridsandGreenITSystems

266

APPENDIX

Table 1: Capability Building Blocks of Energy Efficicent Data Centres.

Level 1 Level 2 Level 3 Level 4 Level 5

Organizational

Structure

No formal

organizational

structure

Resource efficiency is

considered by IT and

facilities, but there

remains a siloed or

disjointed approach.

Resource efficiency is

inherent in policies.

Management of the DC

takes account of the

interrelationship of IT

and Facilities.

Holistic management

approach with decisions

balancing sustainability,

resilience, and business

needs.

A team led by a senior

manager has

responsibility for

Resource Efficiency

across the enterprise.

Policy

No formal

resource efficiency

policies in place

IT policies have limited

consideration for

decommissioning,

consolidation, refresh,

efficient storage

allocation, and

virtualization. Policy

creation is essentially

siloed.

Policy moves towards

increased virtualization.

Facilities have a defined

improvement roadmap

that targets sustainable

operations. External

best practices are

systematically reviewed

and internalised.

Policies reflect a

harmonised,

process-based

approach. Resource

efficiency is a criterion in

offerings and purchases.

There are CapEx

funding programmes for

upgrading infrastructure.

DC resource efficiency

policy is a continuum from

the enterprise level to the

software code level and

everything in between.

Manageability

and Metering

No specific

energy-related

metrics or

metering capability

in place.

Basic information

systems exist for

energy data analysis

and decision support.

IT electrical load

measures at the UPS

level. PUE and DCiE

are used.

The IT organization has

a granular

understanding of its IT

electrical load. Facilities

have an increased level

of the support

infrastructure metered.

IT has rack and server-

level consumption data,

together with

environmental data such

as temperature and

humidity. Facilities

infrastructure is

completely metered from

an energy standpoint.

IT can measure electrical

load at the service level,

matching consumption to

useful work done.

Facilities infrastructure

and IT infrastructure is

completely metered with

appropriate, optimized

automation.

Infrastructure

& Services

Ad hoc

Defined IT landing

procedures consider

resource efficiency.

There is auditing and

decommissioning of

unused equipment.

A comprehensive

consolidation

programme is in place.

DC has moved some

legacy services to

virtualized

environments.

Virtualization is the

default practice for

server and storage

provisioning.

DC is almost exclusively

virtualized. Dynamic

service management

allows for transferable

workloads. IT moves

towards a machine-

readable SLA.

Internal Air &

cooling

Ad hoc design and

operation

IT Equipment is

oriented in a cold

aisle/hot aisle

configuration.

Air inlet supply

temperature is at the

lower end of the

ASHRAE

recommendation. Row-

based cooling maybe

utilised.

Full air segregation is in

place. There is cold

aisle/hot aisle, chimney

cabinets, or in-rack

cooling. CRAC/CRAH

have VFDs.

An optimal, floating, HUM

setpoint is used. DC

normal operational mode

is 'free-cooling'

economization.

Cooling Plant

Cooling is typically

supplied based on

resilience

Refrigeration

infrastructure is

appropriately sized or

strategies are in place

to align cooling

capacity and demand.

COP is typically ~4-6.

Pumps have VFDs.

Partial wet side or

airside economization is

increasingly utilised – for

~50% of the year.

Refrigeration fans and

pumps have VFDs. COP

is typically ~6 or greater

and normal operation is

eco mode for ~75% of

the year. Wet side eco /

evaporation or direct

free air cooling utilised.

Evaporative cooling (wet

side economization), or,

where possible, direct

free air cooling is used.

Direct touch cooling or

new technologies maybe

utilised.

Power

Infrastructure

Power is typically

supplied on a

resilience basis

only.

UPS is more

effectively sized or

strategies are

employed to more

appropriately align

demand and capacity

relative to existing IT

load.

UPS is correctly sized.

UPS is typically ~93%

efficient or above at 50%

load. There is an optimal

number of PDUs.

UPS is correctly sized to

DC load. Redundancy is

appropriate for the

criticality of the load.

Rack PDUs are efficient

with less than 3% loss.

Power is delivered on a

dynamic basis.

UPS is modular and

efficient for the given DC.

Appropriate non-critical

applications are on

mains-only power.

Renewable energy

sources are integrated,

possibly utilizing direct

current in major retro-fit

scenarios.

AMATURITYMODELFORENERGYEFFICIENCYINMATUREDATACENTRES

267