Cost and capacity pressures on the corporate data center are mounting. Increasing computing power demands, poor asset utilization, excess complexity, and growing concerns about energy usage and costs are forcing companies to reassess how they manage their data centers. Companies that don't do so face a future of rising costs and declining performance relative to their competitors. Companies that do make the effort can expect to cut the cost of operating their data centers by as much as 40 percent.
1. Perspective Stefan Stroh
Dr. Germar Schröder
Dr. Florian Gröne
Keeping the Data
Center Competitive
Six Levers for
Boosting Performance,
Reducing Costs, and
Preparing for an
On-Demand World
2. Contact Information
Beirut Frankfurt London New York
Ramez Shehadi Stefan Stroh Louise Fletcher Jeff Tucker
Partner Partner Partner Partner
+961-1-336433 +49-69-97167-423 +44-20-7393-3530 +1-212-551-6653
ramez.shehadi@booz.com stefan.stroh@booz.com louise.fletcher@booz.com jeff.tucker@booz.com
Berlin Dr. Germar Schröder Milan Sydney
Dr. Florian Gröne Principal Enrico Strada Chris Manning
Senior Associate +49-69-97167-426 Partner Partner
+49-30-88705-844 germar.schroeder@booz.com +39-02-72-50-93-00 +61-2-9321-1924
florian.groene@booz.com enrico.strada@booz.com chris.manning@booz.com
Hong Kong
Chicago Edward Tse Mumbai Tokyo
Mike Cooke Senior Partner Suvojoy Sengupta Shigeo Kizaki
Partner +852-3650-6100 Partner Partner
+1-312-578-4639 edward.tse@booz.com +91-22-2287-2001 +81-3-3436-8647
mike.cooke@booz.com suvojoy.sengupta@booz.com shigeo.kizaki@booz.com
Christopher Schmitz and Christian Beekes also contributed to this Perspective.
Booz & Company
3. EXECUTIVE Cost and capacity pressures on the corporate data center are
mounting. Increasing computing power demands, poor asset
SUMMARY
utilization, excess complexity, and growing concerns about
energy usage and costs are forcing companies to reassess how
they manage their data centers. Whether they operate their
data centers for internal customers or as third-party providers
of data center services to others, companies that don’t make the
effort to rethink their data center strategy face a future of rising
costs and declining performance relative to their competitors.
Companies that do make the effort can expect to cut the cost
of operating their data centers by as much as 40 percent.
Managers of data centers should look at six areas in which
their operations can be improved. The greatest potential sav-
ings lie in improved utilization of data center assets, through
server and storage virtualization and by making better use of
the data center facility itself—including a careful analysis of
the total worldwide footprint of data center facilities as well
as how operations are organized. Understanding how and
when data center resources are consumed can further improve
asset utilization and save energy. Restructuring the data cen-
ter’s operating model can increase efficiency, as can devising
a global sourcing strategy for data center services. Finally,
moving to a demand-driven model that rationalizes platforms
and products will help set the stage for the creation of the data
center of the future, one that can give corporate customers
what they want: efficient and flexible computing capacity.
Booz & Company 1
4. Key Findings
• ven as demand for data services is
E
on the rise, the data center is under
tremendous pressure to cut costs,
reduce energy usage, and develop
new delivery models.
• hose pressures, and the threat of
T
rising costs, will force every data
center operator to reassess how it
does business if it wishes to remain
competitive.
• e believe there are six areas in
W
which companies can work to im-
prove their data center operations:
• mprove asset utilization through
I
virtualization Rethinking Deutsche Bank, and DHL—are
making major investments in state-
• onsolidate the data center foot-
C the of-the-art data centers in hopes of
print Data Center making their data center operations
more efficient and less costly. Such
• anage consumption to reduce
M efforts can lower total capital and
usage and energy costs operating expenses by as much as
• estructure data center manage-
R 40 percent. If corporations desire
ment and operating model to maximize the value of their data
At a time when every large-scale orga- center assets and reach the next level
• reate a global data services
C nization is looking to cut expenses and in performance, cost efficiency, and
sourcing strategy streamline operations, the data center quality control, they must begin now
has come under increasing pressure to to rethink the core structures of their
• odularize services offerings and
M
make its operations leaner. And the data center production model.
rationalize payment schemes
time is ripe: Traditional data centers
are facing the upper limits of their Creating the data center of the future
data capacity even as they continue to will require a reassessment of the
underutilize computing assets, while current model in six specific areas:
their massive appetite for electrical technology platforms, data center
power continues to raise concerns topology, consumption management,
about their impact on the environment. end-to-end process efficiency, global
sourcing models, and commercial
Any number of information-intensive models. The risk of not doing so?
companies—including the likes Falling behind in the very competitive
of Google, Microsoft, Facebook, race for IT efficiency.
2 Booz Company
5. Pressures Four factors are driving the need
to rethink the data center. First,
typical data center is saddled with
a large and unwieldy inventory of
on the despite ever more powerful comput- operating systems, database software,
Data Center ing technologies, such as multi-core,
64-bit chip architectures, the installed
and middleware. That in turn adds
hugely to the complexity of systems
base of servers has been growing administration and maintenance—not
12 percent a year, from 14 million to mention adding to cost. And server
in 2000 to 35 million in 2008. Yet utilization is often low, with too many
that growth isn’t keeping up with the CPUs idle or nearly so. Poor utiliza-
demands placed on data centers for tion is a central cause of inflated data
computing power and the amount of center capacity requirements, unnec-
data they can handle. Almost 30 per- essarily high investments in hardware,
cent of respondents to a 2008 survey and big energy bills.
of data center managers said their
centers will have reached their capac- Moreover, companies face grow-
ity limits in three years or sooner. ing concerns about the enormous
amounts of energy their data centers
At the same time, too many data use and the resulting high levels of
centers are just not managed very CO2 emissions they cause; indeed, a
well. Thanks to long histories of number of European countries are
legacy systems and software, a lack looking to regulate emissions even
of discipline regarding standards, and more strictly than they do now.
ineffective life-cycle management, the European data centers currently
European data centers currently
consume more energy than
the entire country of Denmark;
by 2019 they are expected to
use more than the Netherlands.
Booz Company 3
6. consume more energy than the entire tion. The auctioning of emission Exhibit 1), even as both external and
country of Denmark; by 2019 they certificates to utilities, slated to begin internal customers apply downward
are expected to use more than in 2012, will further increase the cost pressure on the price of data center
the Netherlands. A European Code of electricity. services. Much of that increase is
of Conduct is already in place; due to the rising cost of energy and
companies subscribing to the code There is only one future for compa- higher labor costs, including wage
must make a voluntary commitment nies whose data centers continue to inflation, an aging, more highly paid
to reduce power consumption by be traditionally operated: a world workforce, and a coming shortage
applying best practices such as of rapidly rising costs. Under cur- in the skills needed to operate data
energy audits, specific action plans rent practices and assuming constant centers, all of which will result in
for reducing emissions, and continu- capacity, data center costs will rise 17 stagnating productivity. Although
ous monitoring of energy consump- percent over the next four years (see continuing improvements in hardware
Exhibit 1
Over Time, Manpower and Energy Cost Inflation Will Eat Up Traditional Operators’ Margins
Enterprise Computing Cost Outlook
Indexed Cost, Assuming Zero Volume Growth Stable Delivery Model
117
110 112
105 • Wage inflation
100 • Aging workforce
• Skill shortages
50 • Stagnating productivity
45 47
43
41
• Improving hardware performance
• Stable price/performance ratio
5%
36 36 36 • Rising power density
35
35 • Energy prices and emissions trading
• Capacity shortages
10 11 11 12
7
• ncreasing construction
I
17 17 18 18 19 and facility services costs
2008 2009 2010 2011 2012
Data Center Margin
-5%
Manpower Cost
Hardware/Software Cost
“Business as usual”
will turn a 5% profit into a -8% Energy Cost
11% loss within 4 years Facility Cost
-11%
Source: Booz Company Econometric Data Center Planning Model
4 Booz Company
7. performance will mean that the ratio Only by reinventing how they run
of price to performance will remain their centers—rethinking everything
steady, rising energy costs will eat up from technology platforms and data
those gains, as will higher costs of center topology to consumption
construction, operations, and facility management, global sourcing models,
services. The inevitable result: Driven end-to-end process efficiency, and
by bottom-line cost pressures alone, commercial models—can providers
providers of data center services will of data center services hope to thrive
see their already thin margins erode despite the pressures they face (see
further, turning average profits of Exhibit 2). How should data center
5 percent of revenues in 2008 into operators work to transform their
losses of 11 percent by 2012. businesses in each of these six areas?
Exhibit 2
CIOs Must Rethink the Core Structures of Their Production Model to Remain Competitive
Structural Optimization Levers
Commercial Models Technology Platforms
• Capacity-on-demand is the wave 6 1 • y increasing utilization, server
B
of the future, requiring service and storage virtualization
modularization and has the greatest potential for
transparent payment schemes. lowering data center costs.
• Potential gains: N/A • Potential gains: 15%–20%
Global Sourcing Models Next Level … Data Center Topology
• Optimizing the global delivery • Cost Efficiency • ootprint consolidation and
F
strategy involves balancing 5 2 a proper tier structure can save
• Performance
commodity applications • Quality Control money, but a balance must be
with more complex tasks struck between scale and complexity.
• Potential gains: 10%–15% • Potential gains: 15%–30%
E2E Process Efficiency Consumption Management
• Restructuring both the data center • etter utilization of data center assets
B
management organization and the 4 3 and reduced energy consumption
operating model can lead to better capacity offer significant benefits in cost and
planning and lower administrative costs. reduced complexity.
• Potential gains: 5%–15% • Potential gains: 5%–25%
Note: All figures should be read as “total cash-out reduction potential”; they include both capital and operational expense reductions.
Source: Booz Company analysis
Booz Company 5
8. The Virtual Of all the steps that can be taken
to reduce costs, using data center
amounts of operational and capital
expenditures that could be better used
Data Center assets more efficiently has perhaps elsewhere. After a detailed analysis
the greatest potential for generating of its data center costs, one large
significant savings. Underutilization European corporation found that a
of both computing and facility assets large-scale virtualization program had
remains a large problem in data cen- the potential to lower its overall data
ters: Servers typically run at less than center costs by 29 percent.
10 percent of capacity, and it is not
uncommon for more than 50 percent The economics of virtualization
of data center floor space to sit under- are powerful (see Exhibit 3). The
utilized as well. The result: significant technology has the potential to lower
Exhibit 3
Virtualization Technology Has Tremendous Potential to Drive Cost Reduction throughout the Data Center
Virtualization Cost Impact Virtualization Economics
Average Cost per Windows Instance
Illustrative Net Effect/Instance1
€12,600 Virtualization Economics
-29%
Low High
Each full-time employee can manage 60 or more
virtual servers, compared with the current 20 to 30 -30% -45%
5,500 dedicated servers.
Manpower €8,950
Virtualization can more than double the current
dedicated hardware utilization rates of less than 10%, -15% -35%
although cost per CPU will be higher.
3,200
Operating systems, databases, and middleware can be run
virtually, although overall expenses may increase, thanks to +15% -10%
Hardware 3,800 added costs for virtualization management software.
2,900 Each dedicated server currently uses 300 to 500 watts
of electricity, compared with an average of just 100 watts +45% -70%
per virtual server.
Software 1,900
2,100 Each dedicated server currently needs, on average,
Energy 800 one to two rack units; that can be reduced to less than -25% -50%
300
Data Center Facility 600 450 one rack unit per virtual server.
Instance on Instance on
dedicated server virtual server
1
Size of effect depends on ratio of dedicated servers to newly virtualized servers, type of virtualization platform, degree of standardization, power density, and tier level of data center.
Source: Booz Company analysis
6 Booz Company
9. total costs of ownership by as much tion—primarily high-volume platforms includes, on the hardware front, an
as 40 percent. Consider hardware. such as Windows and standard Unix inventory of assets—the number
Dedicated servers frequently have or Linux—lies somewhere between 20 of harmonized vendor clusters and
utilization rates of less than 10 and 60 percent of assets, depending on machines with fewer than four
percent, whereas servers run virtually the production model and computing CPUs—and an analysis of utilization
can often more than double those footprint. Trying to virtualize more levels. As to software, the inventory
rates. And even though virtual servers than that will usually involve the virtu- should include harmonized operating
typically cost more per CPU, the alization of more exotic platforms with systems, database software and
overall benefit can be hardware cost lower server counts, such as legacy middleware clusters, and standard,
savings of between 15 and 35 percent. Unix systems, and the effort simply multi-platform certified applications
won’t bring the returns expected. such as Web and e-mail and standard
Despite the very real benefits of virtu- ERP and CRM. Finally, it’s important
alization, it rarely makes sense to try to Achieving the maximum return to ascertain whether any applications
virtualize everything. Indeed, the “sweet requires a careful review of the have technical restrictions, such as
spot” for generating the maximum preconditions in the data center maintenance liabilities, that might
return on investments in virtualiza- for successful virtualization. That restrict the use of virtualization.
Dedicated servers frequently have
utilization rates of less than 10 percent,
whereas servers run virtually can
often more than double those rates.
Booz Company 7
10. Mapping the Large multinational corporations use
different strategies for siting their data
ated by excess complexity have
not begun to make themselves felt
Data Center centers. Some may find themselves (see Exhibit 4).
running dozens of data centers around
the world. Hewlett-Packard, for Another way to put the problem is
instance, maintains about 60 centers in terms of utilization. Obviously,
worldwide. Others, such as ING, data centers are expensive. Looked
maintain just one primary hub. Data at in terms of cost as a function of
center topology, however, creates a utilization rate, however, unit costs
dilemma: Scale or resilience? A topol- come down rapidly. But the benefits
ogy that includes a small number of go only so far. After about 90 percent
large-scale centers offers the benefit utilization, data centers run a real
of scale, but the lack of diversification risk of losing operational flexibility.
can pose a security risk, and individ- Generally speaking, the utilization
ual centers risk being simply too com- goal should be about 80 percent,
plex to operate efficiently. A plan that which leaves adequate headroom for
includes many smaller centers runs the peak demand. Service providers will
opposite risk: Security concerns and want to leave somewhat more room
complexity are eased, but the individ- depending on their mid-term deal
ual centers may not be large enough pipeline, which may add sudden large
to reap the maximum benefits of scale. demands on their data centers. By the
Where is the happy medium? same token, in coping with capacity
bottlenecks, before expanding capac-
To operate at peak efficiency, data ity in a lower-tier center and reducing
centers should be about 10,000 its utilization rate, consider the pos-
square meters. At that size, each sibility of using higher-tier capacity
center’s annual operating expenses are with better utilization and overall
minimized, but the added costs cre- lower unit costs.
Exhibit 4
When Consolidating Data Centers, It Is Vital to Find the Right Balance between Scale and Complexity
Data Center Cost by Size
16,000
Annual Operational Expense (€ per Square Meter)
Efficient Data
Case A
Center Size
14,000
Case B
Case C
12,000
Case D
10,000 Case E
8,000
6,000
4,000
2,000
0
0 5,000 10,000 15,000 20,000
Scale Effects Dominate Complexity Costs Erode Scale Effects
Computing Floor Space (in Square Meters)
Source: Booz Company analysis
8 Booz Company
11. Managing On the other side of the utilization
coin is the issue of consumption man-
changes in resource utilization in order
to understand why they occur. These
Consumption agement, involving the consumption steps can reduce consumption of assets
of both computing assets and energy. by up to 15 percent.
Significant savings can be found in the
effort to reduce the use of assets and A further 15 to 20 percent reduction
to optimize the kinds and number of in the use of resources can be achieved
assets being used. The key here is to by identifying, and balancing, differ-
implement an efficient capacity plan- ences in utilization by region, season,
ning process. On the consumption time of day, and line of business.
side, begin by identifying the resources Work with application owners and
required for each software application. application development teams to
Then retire or move any resources identify the factors driving utilization
that do not get accessed frequently. and to develop measures for reducing
Together with the application develop- consumption, including the renegotia-
ment team, work to limit increases in tion of service-level agreements, the
consumption that may occur when restructuring of job networks, and the
new application releases are rolled out. redesign of applications to run at peak
Set up a program to closely monitor efficiency. Again, devise a program
A further 15 to 20 percent reduction
in the use of resources can also be
achieved by identifying, and balancing,
differences in utilization by region,
season, time of day, and line of business.
Booz Company 9
12. to monitor utilization and how the the CPUs consume the most energy. number of AC-DC/DC-AC conversion
measures you have taken are affecting Ideas for reducing the amount of cycles and by converting to high-
resource utilization rates. power consumed by CPUs include efficiency power distribution systems.
virtualization and the use of more Cooling costs can be lowered by the
Data center operators can take a efficient multi-core processors and use of district cooling, heat pumps to
variety of steps to save money on processors with dynamic scaling. reduce fan loading, desiccant cooling
the energy side (see Exhibit 5). Of Gains can also be made in the area of driven from waste heat, variable
the various data center components, power distribution by reducing the speed fans, and direct liquid cooling.
Exhibit 5
Data Center Operators Can Deploy a Number of Effective Measures to Optimize Energy Consumption
Typical Data Center Energy Usage by Component (%)
100
• igh-efficiency systems (e.g., multi-core
H Processor load
90 processors, virtualization, processors with Power system load
dynamic frequency scaling, silicon storage, etc.
Cooling system load
80 41% 41%
Potential improvement measures
70 • educed AC-DC/DC-AC conversion cycles
R
% Power Consumed
• High-efficiency power distribution
60
• District cooling
50 20% • Heat pumps to reduce fan loading
• Desiccant cooling driven from waste heat
40 37% • Variable speed fans
12% • Direct liquid cooling
30
8%
20
7%
10 23% 6%
2%
4% 1%
0
Total CPUs Power Chillers Uninterruptible Voltage Server Fans Computer Power Water
Supply Units Power Supply Regulators Room Fans Distribution Pumps
Component
Source: Data Center Energy Briefing, U.S. Department of Energy; Intel Corporation; Booz Company analysis
10 Booz Company
13. The Efficient The typical data center faces a further
challenge: the lack of truly efficient
Many of the causes of inefficiency can
be attributed to organizational prob-
Data Center organizational structures and pro- lems such as understaffed demand
cesses. The causes of inefficiency and capacity planning functions and
are many: Too many data centers the lack of an integrated operating
find themselves focusing on day-to- model. After a careful analysis of its
day troubleshooting rather than on employees’ activities, one company
strong system architecture and design. running a midrange hosting operation
Furthermore, data centers often move discovered that employees were spend-
into production mode prematurely, ing far too little time on planning and
before they have completed proper building out their systems, and far too
testing and deployment procedures. much time on daily operations and
The result is a low degree of stan- ad-hoc troubleshooting. The result:
dardization in commodity operations Day-to-day operations struggled with
activities and no clearly modularized a poorly integrated operating model—
service and product portfolios, which and efforts to standardize infrastruc-
makes both sales and product manage- ture and increase utilization were
ment needlessly complex. Incoherent doomed from the start.
process routines and lack of fully
transparent end-to-end service man- The consequences of a poorly orga-
agement often lead to the delivery of nized operating model can be dire
service levels over and above what has (see Exhibit 6). In the model on the
been agreed to (and is being paid for) left, the various functions of the data
by the customer—24/7 support, for center, from hosting to storage to con-
instance, becomes the default setting. nectivity, are effectively siloed, with
Exhibit 6
A Cross-Platform Planning and Management Capability Can Improve Efficiency
Typical Data Center Operating Model Integrated Data Center Operating Model
Customer-Facing Functions Customer-Facing Functions
Capacity Planning Administration
Admin. Admin. Reduce Costs
Admin. Admin.
Risks
Service Service Management
Manage- Service Service Service
ment Manage- Manage- Manage-
ment ment ment
Service Service Service Service Service
Service Service Service Delivery Delivery Delivery Delivery
Delivery
Delivery Delivery Delivery
Host Midrange Storage Connectivity Host Midrange Storage Connectivity
• urrent data center management is fragmented into many layers,
C • By integrating service management and thinking in terms of
with too many handoffs of core processes, such as problem resolution “services,” not “servers,” data centers can achieve better capacity
and operations and change management, between functions. planning and management and lower administration costs.
Source: Booz Company client example
Booz Company 11
14. each function running its own admin- with the goal of automating routine, features can help with compliance
istration and service management and labor-intensive tasks such as trouble and risk management tasks, and some
delivery. The resulting fragmentation ticketing, fault management, and suites offer the ability to manage
creates the need for an excessive performance management. And the workflow aligned with standard ITIL
number of handoffs when problems number of automation tools is growing processes. The maturity of such suites
occur, and any effort of the various fast, as are the different configurations remains a concern, however. Many of
functions to work together to change of these automation systems, and the them still lack real depth and interop-
operating procedures becomes very move to virtualization will only add to erability, and they do not typically
difficult. Instead, planning, admin- that complexity. As long as each plat- cover critical areas such as storage area
istration, and service management form possesses its own management networks and other network functions.
should be integrated across all the silo, moreover, each silo will look to The market includes a number of niche
functions, as in the model on the right, automate its own platform operations. players in such areas as server provi-
allowing for better capacity planning sioning, migration to virtual machines,
and lower administration costs. Vendors are offering a variety of patch management, and storage
automation suites that can aid in the allocation. The result: Automation
Can the automation of data centers process of automation. Such tools efforts still require a patchwork of
help improve efficiency? That, of already include configuration manage- tools—BMC Patrol combined with
course, is the hope of every operator of ment functions such as automatic asset MS System Center, for instance, or
data centers, especially as both process discovery and resource transparency VMware vCenter Server and EMC
and management complexity increases. and are beginning to offer dependency ControlCenter. Buyer, beware.
The desire to raise the efficiency of mapping and advanced configuration
IT processes themselves is strong, item reporting. New audit and control
One company running a midrange
hosting operation discovered that employees
were spending far too little time on
planning and building out their systems,
and far too much time on daily
operations and ad-hoc troubleshooting.
12 Booz Company
15. Global As corporations look to rationalize
and save money on their overall data
ers—continues to grow quickly. A
well-planned and well-executed global
Delivery center footprints, the opportunity to delivery model for data center services
offshore and nearshore a variety of can generate considerable savings,
data center services—and to farm out depending on which services are sent
some services to third-party provid- offshore, where they are sent, and
to whom. Offshoring or nearshoring
suitable “commodity” activities such
as application management, database
management, monitoring, and engi-
Exhibit 7 neering will bring the greatest cost
The Market for Offshore and Nearshore Services Continues to Gain Momentum, decreases—thanks primarily to lower
Offering Significant Cost Reduction Opportunities
salary and benefits costs, significant
process improvements, and lower
Offshore/Nearshore IT Production: Examples overall management costs.
Total Cost Structure Shifts
Application management Midrange server operations Moving application management
-22% -10%
efforts to a combined onshore/
nearshore model probably offers the
Manpower 37%
Manpower 69%
29% greatest benefit—cost reductions of
56% up to 22 percent—primarily because
Hardware Hardware 28% 31%
Software 0%
1% of the large reduction in labor costs.
27% Software 11% 12%
Connectivity
21%
3% Data Center 14% 16%
Manpower typically makes up fully
0%
Other 10% 13% Other 10% 11% 69 percent of the cost of onshore
Onshore Onshore/Nearshore Onshore RIM Nearshore
application management; moving to a
combined
combined model, however, can reduce
Mainframe operations Storage operations
-6% -6%
labor costs to just 48 percent of the
total. Cost savings can also be found
Manpower 16% 13%
Manpower 30% 28% in other remote management models,
Hardware 19% 19% Hardware 43% 44% including mainframes, infrastructure,
and storage (see Exhibit 7).
Software 28% 29% Software 14% 14%
Data Center 11% 12% Data Center 15% 15%
Other 12% 12% Other 13% 13%
Onshore RIM Nearshore Onshore RIM Nearshore
RIM=Remote Infrastructure Management
Source: Booz Company analysis
Booz Company 13
16. Companies have taken a variety of as eastern Europe gain expertise in A third company, a large European
routes in their efforts to reap the running large-scale data centers. industrial manufacturer, outsourced
cost and flexibility benefits of global Looking to create a shared onshore/ all of its data center operations,
sourcing. India has long been a prime offshore infrastructure delivery including remote management
region for such activities. One large model, one German corporation of more than 5,000 servers, to a
data center operator, for example, recently set up a nearshore subsidiary provider in India. In addition to data
turned to an Indian outsourcer for to provide its more commoditized services, the scope of the project
services that included 350 servers, 46 second- and third-line services on a included incident management,
separate databases, 3,200 network 24/7 basis, in addition to its onshore monitoring, and change execution,
elements, and 10 firewalls. In data center, where the tasks requiring all provided on a 24/7 basis. The
addition, the provider offered virus a more highly skilled workforce project began with just 50 full-time
management, as well as backup and took place. Again, the benefits came employees, but within three years that
storage management. The benefits primarily in the form of lower labor number was approaching 150. Again,
obtained included service levels costs. Wages were about 40 percent the advantage in labor costs provided
of 99.95 percent, 24 hours a day, lower than those in Germany, and the greatest savings: Wages in India
and higher productivity, as well as the average age of the nearshore averaged less than half of those in
significant wage differentials. employees was likewise lower. The the European headquarters. And
client also found a strong talent pool the project also allowed the client
Such benefits can be found closer surrounding its new facility, with to enforce a high level of process
to home as well, as regions such adequate English and IT skills, thanks standardization and automation.
to the presence of nearby universities.
Manpower typically makes up fully
69 percent of the cost of onshore application
management; moving to a combined
onshore/nearshore model, however, can reduce
labor costs to just 48 percent of the total.
14 Booz Company
17. Managing Data center operators looking to
offer their services to others on a
and the revenues obtained from them.
Without that connection, they cannot
the Third- commercial basis frequently face the clearly link capacity planning and
Party same problem: Too often they don’t
have the ability to conduct effective
revenue projections, and they cannot
determine clear targets for their costs
Data Center end-to-end capacity management of production, given what the market
and cost steering, or the means to expects. Solving this problem requires
provide transparency for costs and simple, transparent connections
services. Typically, their catalog between services and modular
of services is too diversified, with production building blocks and the
every customer enjoying its own set capacity to understand the market
of specific, personally configured and link it to target cost planning.
services. The result is an overly
complex set of offerings that is A further consequence of the typical
difficult to benchmark or rationally collection of diversified services
and transparently charge for. Making maintained by most commercial
the transition to a demand-driven data centers is a lack of transparency
model will require significant changes regarding the total cost of the
in how data centers operate. They data center’s operations and poor
must move to a standardized set of governance in managing those
limited platform products that can costs. The mechanisms by which
be strictly managed and maintained, costs are allocated become overly
easily compared with the offerings of complex, thanks in part to poor
competitors, and straightforward to organizational alignment. Here again,
cost out. the solution lies in introducing a cost-
reporting mechanism that is simple,
In the area where service and transparent, and based on total cost
production issues merge, the problem of ownership, and in realigning
is similar. Many data centers struggle the organization into “production
to maintain a clear, logical connection towers” that can help organize the
between how services are produced process of cost reporting.
Booz Company 15
18. The capacity-on-demand model customer’s needs, and based more on
is clearly the future of data center computing capacity and performance
operations (see Exhibit 8). Customers than on specific hardware and
no longer want to pay for capacity software configurations. That
they aren’t always using, and they means developing efficient new
don’t much care anymore about the ways to deploy capacity and shut
specifics of the hardware and software it down when not needed, to better
being employed. In order to stay balance total capacity usage, and
competitive, providers must move to standardize both hardware and
quickly to offer data center services software platforms so they can scale
that are scalable depending on the up capacity quickly.
In Summary
These six areas in which data center
operations can be enhanced offer
data centers the potential for major
improvements in their performance,
Exhibit 8 and significant benefits in the form
The Continued Move toward Capacity-on-Demand Models Will Force of reduced operating costs. Both
Data Centers to Rationalize Platforms and Services Offerings corporate data centers and centers
providing computing services to
Data Center Service Model Trends others should consider some or all of
the improvements suggested if they
wish to maintain their competitive
position. The data center is rapidly
Classical
Outsourcing moving toward a new model in which
what matters is delivering as much
computing capacity as customers
Spend Managed
Services
need, when they need it. Will you be
ready to give it to them?
Computing
Utility
0
2005 2020
Servers Toward variable commercial delivery Services
• Platform specific • Platform independent
• ased on individual system
B • ased on computing capacity
B
configurations customized and performance
for each customer • calable depending on needs
S
Source: Booz Company analysis
16 Booz Company
19. About the Authors
Stefan Stroh is a partner with Dr. Florian Gröne is
Booz Company in Frankfurt. a senior associate with
He leads the global transporta- Booz Company in Berlin.
tion technology practice and He supports telecommunica-
works for leading players in the tions companies and ICT
international railway, logistics, Service Providers in develop-
aviation, travel, high tech, and ing their market positioning
consumer products sectors. strategies and improving IT
operations efficiency. He also
Dr. Germar Schröder is a works on CRM strategy and
principal with Booz Company architecture across industries.
in Frankfurt. He focuses on
IT strategy, large-scale transfor-
mation programs, and finance
IT, primarily for the telecom-
munications industry. He also
supports IT service providers
in business model develop-
ment, strategy, and operational
efficiency.
Booz Company 17