Académique Documents
Professionnel Documents
Culture Documents
www.emeraldinsight.com/1463-7154.htm
BPMJ
17,3
510
Tugrul U. Daim
Portland State University, Portland, Oregon, USA
Mitali Monalisa
Intel Corp, Hillsboro, Oregon, USA, and
Jay Justice
Motorola Mobility, Beaverton, Oregon, USA
Abstract
Purpose This paper aims to research literature to describe the business processes used when
planning IT infrastructure refreshes.
Design/methodology/approach The paper uses analytical hierarchical process (AHP) and
pairwise comparisons to model and quantify the decision process for IT infrastructure refreshes.
Findings The research found that most companies keep their refresh processes private and very
little academic research is available on this topic. While supportability, manageability, compatibility,
cost, and scalability are important factors to large organizations, performance and availability of the
systems are important for smaller organizations.
Originality/value AHP was not ever used to evaluate the refresh planning. The paper
demonstrates that it would be a very useful tool.
Keywords Business process re-engineering, Information technology, Analytical hierarchy process
Paper type Literature review
Introduction
The rapid growth of computing technologies such as cloud computing and
virtualization is driving companies to refresh their IT infrastructure, sometimes
sooner than planned, in order to support these new technologies (Gartner Inc., 2009). In
mid to large enterprises with a complex IT infrastructure, a refresh can be costly and
require careful planning in order to get the largest return on investment over time. The
authors of this paper researched literature to describe the business processes used
when planning IT infrastructure refreshes. Our research found that most companies
keep their refresh processes private and very little academic research is available on
this topic.
This paper uses the analytical hierarchical process (AHP) decision-making model
to help companies plan their IT infrastructure refresh. The model is intended to have the
flexibility for any IT organization regardless of its size or strategic alignment. The model
uses pairwise comparison in order to tailor the criteria to the specific IT organization.
When implemented, the model will recommend one of seven possible strategies to
refresh IT infrastructure.
Literature review
There are two important aspects to this paper the AHP model and applying it to an IT
infrastructure refresh. In researching these two areas, we found plenty of literature on
decision-making models, specifically AHP, but very little academic research has been
conducted on the processes used by businesses to update their IT hardware. Our
literature review will start with an overview of AHP and conclude with the research
found on refreshing IT infrastructures.
Saaty and Kearns describe the AHP as a decision-making framework for dealing
with problems with multiple criteria to be considered (Saaty and Kearns, 1985).
The AHP model consists of four stages (Roper-Lowe and Sharp, 1990):
.
The first stage, building the decision hierarchy, is the most important stage as it
sets the framework for the rest of the process. At the highest level is the goal to
accomplish. To reach that goal, a number of criteria need to be met; these are
represented in the second tier. If necessary, the criteria can be further divided into
sub-criteria. At the bottom of the hierarchy are the possible alternatives to be
considered to meet the goal at the top.
.
In the second stage, values are assigned to each criterion. This weighting can be
done one of two ways using quantitative engineering requirements or using a
qualitative method that can turn preferences into values such as pairwise
comparison (Kocaoglu, 1983).
.
The third stage is scoring each alternative with respect to each criterion. This
requires finding people qualified to make judgments, i.e. a finance person could
score the alternatives with relation to cost but an engineer would be needed to
judge the technical specifications (Roper-Lowe and Sharp, 1990).
.
The final stage is applying the weights to the alternatives to obtain an overall
score and determine the preferred option to reach the goal.
In stages two and three above, a technique called pairwise comparison is used to find the
weighting of criteria and alternatives. There are several models that can be used to do the
comparisons but typically an expert panel is used. The expert panel method is
commonly used when there are intangible criteria to be considered in the decision
(Roper-Lowe and Sharp, 1990). One of the disadvantages of using the weights derived
from pairwise comparison is that it assumes a linear relationship between the value and
the user preferences. Gerdsri and Kocaoglu (2007) suggest using desirability curves that
present a higher resolution of the users preferences could solve this problem. However,
creating the curves requires a detailed dataset for each criterion and in the interest of
time linear weights were used in this model.
Interactions between criteria and sub-criteria can change depending on the
application of the AHP. In some situations, it makes sense to weigh all criteria against
each other but in others, direct comparisons would not make sense. In order to
accommodate this, AHP models have the flexibility to handle both situations. Gerdsri
and Kocaoglu (2007) show in Figure 1 the application of a three-level decision model
where criteria are selected such that they are independent of each other and have no
IT infrastructure
refresh planning
511
BPMJ
17,3
Objective
(O)
Criteria
(Cb)
512
C1
Factors
(Fnk)
F21
F11
F13
Ck
F23
F31
F12
Fjk1
Figure 1.
AHP model with
independent criteria
C3
C2
Technologies
(Tn)
T1
F22
T2
F32
F1
Fjk2
T3
F2
FjkK
TN
interaction with another group. This simplification makes the model very practical to
use as judgment quantification experts are not asked to compare two sub-criteria that
belong to different groups and difficult to evaluate.
AHP is a very effective and robust tool as evidenced by its widespread use (Chen
and Kocaoglu, 2008). The following list is provided as a sample of AHP applications:
.
Sevkli et al. (2008) supplier selection.
.
Hafeez and Essmail (2007) evaluating core competences.
.
Bhagwat and Sharma (2007) performance measurement.
.
Al-Subhi and Al-Harbi (2001) project management.
.
Karami (2006) adoption of irrigation methods.
.
Leung et al. (1998) evaluating fisheries management.
.
Wong and Li (2008) intelligent building systems.
In researching the strategies companies use to refresh their IT infrastructure,
specifically server systems, there are not many academic studies available. A common
theme in the articles was a need for better planning in regards to IT. Management finds it
difficult to devote resources to improving infrastructure because it is hard to correlate to
business needs and often considered immeasurable (Duncan, 1995). Weill (1992)
presented studies that show a correlation between an increase in business performance
and investment in IT infrastructure; meaning that investment in IT could be a critical
factor in competing with other companies.
However, spending resources on new IT infrastructure to keep a competitive
advantage without a plan could lead to high costs in the future. Brill (2007) found that
less than 20 percent of 100 data center operators surveyed felt that their infrastructure
planning procedures were above average. In his report, Brill found that the
infrastructure costs of power, cooling, and space will exceed the initial cost of the IT
hardware in a three-year period. This means that any new IT deployments that only look
at the cost of hardware could be completely inaccurate. One of the solutions Brill (2007)
suggests is virtualizing older, inefficient hardware onto a single more powerful server,
thereby reducing the power requirements.
Our research showed that the AHP is a useful tool when making decisions with
tangible and intangible criteria to consider. In order to keep the intangible and tangible
factors separate from each other, we will create a four-level decision hierarchy with
independent criteria and use pairwise comparison to obtain the weighting of our criteria.
We have also shown that implementing an efficient and flexible IT infrastructure
can be a strategic advantage over competitors but there are many factors to consider in
order to get the best return on investment.
Methodology
This section explains the research stages used in this paper. The first step was to
develop the hierarchy. A set of industry experts was used for this. Then, in parallel
stages, the model was applied to two cases.
Expert panel
An expert panel made up of the authors was used to determine the criteria and
sub-criteria required to choose a strategy to refresh the servers in an IT infrastructure.
The expert panel consisted of seven individuals from the high-tech industry with
experience ranging from five to 27 years. The job roles of the experts included: senior
database engineers, enterprise technical marketing, and validation engineers. The
choices of criteria and pairwise comparisons only had small inconsistencies between
the experts and were deemed negligible. If the inconsistencies were large then a more
advanced method such as Delphi would have been used to reach a consensus.
Decision hierarchy
The primary requirement used when creating the hierarchy was to keep it flexible
enough to be applicable from small to large businesses. Based on experiences from the
expert panel, we found that, regardless of size, businesses consider the same criteria
when updating their IT infrastructure. Two of the experts used were involved in sever
refresh decision making in the past and helped validate our criteria.
Once the criteria were defined, the expert panel identified possible strategies to
evaluate. Determining the possible alternatives for the lower level of the hierarchy
turned out to be the most difficult part. There are too many possible strategies to
refresh servers so the expert panel narrowed the scope to seven possible alternatives
that would represent the majority of the strategies.
Application
The decision hierarchy should be applied on a regular basis to determine is a server
refresh would have a positive return on investment and, if so, what strategy would best
align with company preferences. In order to apply the model, weights representing the
companys preferences need to be applied to the criteria. One of the ways to get a weight
is to take a survey of employees in relevant decision-making positions. In a small
business, this could be an IT manager and database administrator while in a large
company there could be multiple decision makers from groups such as management,
engineering, or finance. In order to reduce the time needed to compile the results of a
survey, it is recommended to create an electronic version so the results can be
IT infrastructure
refresh planning
513
BPMJ
17,3
Model development
As discussed previously, the decision hierarchy contains four levels, the goal, criteria,
sub-criteria, and the seven alternative strategies. Starting at the top is the goal to
choose a server refresh strategy. All pairwise comparisons of the criteria and
alternatives are done with respect to this goal.
The second level of the hierarchy is the criteria. Each of the criteria in this level
requires a weight that calculates into scoring the alternatives. The values of these
weights are found through pairwise comparisons of each criterion by an expert panel in
the company considering a server refresh. The four criteria identified in this model are:
reliability availability serviceability (RAS), performance, total cost of ownership (TCO),
and roadmap.
The third level in the hierarchy contains the sub-criteria for each criterion. The
sub-criteria under each criterion are independent from sub-criteria under another
criterion, i.e. the sub-criteria under performance cannot be compared to the sub-criteria
under roadmap. The total contribution of any factor for any alternative is calculated by
the product of the criterion weight and the sub-criterion weight. For example, if the
pairwise comparison at the criteria level for RAS indicates a 20 percent weight and the
pairwise comparison for serviceability indicates a 30 percent weight, the final weight
for serviceability will be 20 *30 percent 6 percent. Descriptions of the sub-criteria can
be found in Appendix 1. The complete model is shown in Figure 2.
The final layer of the model is the actions which include the possible strategic
alternatives. In order to derive the preference scores for each alternativem several
assumptions were made. In a real-world application of this model, the actual specifications
Figure 2.
Server refresh
decision hierarchy
Actions
Sub-criteria
Criteria
Goal
514
calculated automatically. This also gives participants instant feedback on how their
responses compare to others similar to the Delphi model. Once the weighting has been
applied, the company can evaluate the possible alternatives and determine which the
preferred option is to refresh their servers.
Reliability availability
serviceability (RAS)
Serviceability
Reliability
Manageability
Availability
Full
replacement
Hardware
upgrade
Performance
Performance/
watt
Transactions
Total cost of
ownership
(TCO)
Recurring
cost
Initial cost
Roadmap
Scalability
Legacy
support
Service life
Upgradability
Density
Partial
replacement
New addition
Virtualization
Software
upgrade
Delay refresh
can be found through research such as benchmarks or retail pricing. The following are the
assumptions made in this paper in order to test the model:
.
virtualization has not yet been implemented;
.
current servers have one year left of service life, while any new servers will have
a three-year service life;
.
new servers are twice as energy efficient as old servers;
.
new servers have four times the processing power and memory as old servers;
.
no previous software or hardware upgrades have been performed on the existing
servers; and
.
the new servers will only take up half the rack space as the old servers.
Based on the assumptions above, the following seven strategic alternatives were identified:
(1) Full replacement. Replace all existing servers with new hardware.
(2) Partial replacement. Replace about 50 percent of the existing servers.
(3) Hardware upgrades. Replace and/or add additional components in the
existing servers. Assumption is any addition will only increase performance
by 20 percent.
(4) New addition. Addition of 10 percent more servers to the existing infrastructure.
(5) Software upgrade. Installation of operating systems and/or applications that
lead to a more efficient computing environment.
(6) Virtualization. Consolidate several servers onto one server using virtual machines.
This model assumes a consolidation ratio of five old servers to one new server.
(7) Delay refresh. Do nothing because the value added of the other alternatives does
not outweigh the upfront cost factor.
Each of the alternatives has a score from zero to one for each of the sub-criteria. For
example, the delaying the refresh alternative could have a very high preference score
for upfront cost since initial cost is zero, but will have lower preference scores for
performance since newer servers will outperform the current hardware. In order to
score each alternative, they are compared to the sub-criteria. If the alternative is the best
option for a given sub-criterion, it is given a score of one. The other six alternatives are
weighed against the best to derive a value that is some percentage from zero to one.
Tangible sub-criteria can be found through specification research. Expert opinion can be
used when insufficient data are available to support specification comparisons.
After all the pairwise comparisons for the criteria are calculated and the preferences
for the alternatives have been identified, a score can be calculated for each alternative
by multiplying the alternative score by each sub-criteria weight and summing the
products. The alternative with the highest sum is an indication which strategy is most
in line with the company preferences.
Applying the model: small vs large businesses
We conducted two case studies to verify our model one large IT environment and
one small IT environment. Two case studies were done to compare and contrast the
differences and determine if the model could work in different situations.
IT infrastructure
refresh planning
515
BPMJ
17,3
516
Table I.
Preference scores for
seven alternatives
Table II.
Case study results
F1 servicability
F2 reliability
F3 manageability
F4 availability
F1 energy efficiency
F2 transactian/sec
F3 density
Fl recurring cost
FZ upfront cost
Fl scalability
F2 legacy
compatibility
F3 upgradability
F4 service life
Virtualization
Full replacement
Partial replacement
Software upgrade
New addition
Delay refresh
Hardware upgrade
1.00
1.00
0.75
0.90
0.20
1.00
0.40
0.80
0.00
1.00
0.25
0.50
0.25
0.85
0.05
0.30
0.20
0.50
0.60
0.30
0.63
0.63
0.15
0.82
0.15
0.63
0.30
0.65
0.50
0.60
0.32
0.32
0.25
0.85
0.10
0.32
0.15
0.53
0.90
0.27
0.45
0.50
0.75
0.80
0.10
0.30
0.30
0.50
0.80
0.30
1.00
1.00
1.00
1.00
1.00
0.25
1.00
1.00
0.80
0.30
0.25
0.25
0.25
0.80
0.10
0.25
0.20
0.50
1.00
0.20
0.30
1.00
1.00
0.90
0.10
0.20
0.50
0.60
0.60
0.80
0.27
0.27
0.50
0.10
0.30
0.20
1.00
1.00
1.00
0.20
0.20
Small environment
Large environment
0.26
0.20
0.15
0.108
0.106
0.105
0.096
0.21
0.20
0.15
0.12
0.113
0.110
0.10
IT infrastructure
refresh planning
517
BPMJ
17,3
518
Brill, K.G. (2007), The Invisible Crisis in the Data Center: The Economic Meltdown of Moores
Law, Uptime Institute, Santa Fe, NM.
Chen, H. and Kocaoglu, D.F. (2008), A sensitivity analysis algorithm for hierarchical
decision models, European Journal of Operational Research, Vol. 185 No. 1, pp. 266-88.
Duncan, N.B. (1995), Capturing flexibility of information technology infrastructure: a study of
resource characteristics and their measure, Journal of Management Information Systems,
Vol. 12 No. 2, pp. 37-57.
Gartner, Inc. (2009), Gartner Identifies the Top 10 Strategic Technologies for 2010, Gartner,
Orlando, FL.
Gerdsri, N. and Kocaoglu, D.F. (2007), Applying the analytic hierarchy process (AHP) to build a
strategic framework for technology roadmapping, Mathematical and Computer
Modelling, Vol. 46 Nos 7/8, pp. 1071-80.
Hafeez, K. and Essmail, E.A. (2007), Evaluating organisation core competences and associated
personal competencies using analytical hierarchy process, Management Research News,
Vol. 30 No. 8, pp. 530-47.
Karami, E. (2006), Appropriateness of farmers adoption of irrigation methods: the application of
the AHP model, Agricultural Systems, Vol. 87 No. 1, pp. 101-19.
Kocaoglu, D.F. (1983), A participative approach to program evaluation, IEEE Transactions on
Engineering Management, Vol. 30 No. 3, pp. 37-44.
Leung, P., Muraoka, J., Nakamoto, S.T. and Pooley, S. (1998), Evaluating fisheries management
options in Hawaii using analytic hierarchy process (AHP), Fisheries Research, Vol. 36
Nos 2/3, pp. 171-83.
Vaidya, O.S. and Kumar, S. (2006), Analytic hierarchy process: an overview application,
European Journal of Operational Research, Vol. 169, No. 1, pp. 1-29.
Roper-Lowe, G.C. and Sharp, J.A. (1990), The analytic process and its application to an
information technology decision, Journal of Operational Research Society, Vol. 41,
pp. 49-59.
Saaty, T.L. and Kearns, K.P. (1985), Analytical Hierarchy Process, Pergamom, New York, NY.
Sevkli, M., Koh, S.C.L., Zaim, S., Demirbag, M. and Tatoglu, E. (2008), Hybrid analytical
hierarchy process model for supplier selection, Industrial Management & Data Systems,
Vol. 108 No. 1, pp. 122-42.
Weill, P. (1992), The relationship between investment in information technology and firm
performance: a study of valve manufacturing sector, Information Systems Research,
Vol. 3, pp. 307-58.
Wong, J.K.W. and Li, H. (2008), Application of the analytic hierarchy process (AHP) in
multi-criteria analysis of the selection of intelligent building systems, Building and
Environment, Vol. 43 No. 1, pp. 108-25.
IT infrastructure
refresh planning
519
Criteria
Factors
Serviceability
Reliability
Manageability
Availability
Performance
Energy Efficiency
Transaction/sec
Density
Recurring Costs
Upfront costs
Roadmap
Scalability
Legacy Compatibility
Upgradeability
Service Life
Factor
Description
Serviceability
Reliability
Manageability
Availability
The ability for a failure to be serviced while the server remains operational
The history of failure and perceived future failures of the server.
The ease of supporting the server hardware and software in the companys IT infrastructure.
The percentage of uptime designed into the server implementation
Energy Efficiency
Transaction/sec
Density
Recurring Costs
Upfront costs
Scalability
Legacy
Compatibility
Upgradeability
The cost of incurred over the life of the server. Including energy used, cooling costs,
maintenance, licensing, ROI.
The cost of new server hardware, the loss of depreciation, the installation labor for the new
hardware and software
The ability to incrementally add one or more systems to an existing cluster when the overall
load of the cluster exceeds its capabilities
The servers ability to integrate with previous software and applications used in the computing
environment.
The servers ability to have its components replaced or additional components added to increase
its capabilities
The serviceability life of the server hardware or the application it runs on until it reaches its end
of service life (EOL).
(continued)
BPMJ
17,3
520
RAS
60
PERFORMANCE
40
CRITERIA COMPARISON
RAS
PERFORMANCE
ROADMAP
RAS
TCO
RAS
TCO
PERFORMANCE
PERFORMANCE
ROADMAP
TCO
ROADMAP
FACTOR COMPARISON
RAS CRITERIA
Reliability
Availability
Serviceability
Manageability
Reliability
Serviceability
Reliability
Manageability
Availability
Serviceability
Availability
Manageability
PERFORMANCE CRITERIA
(continued)
Energy efficiency
Transactions/sec
Transactions/sec
Density
Energy efficiency
IT infrastructure
refresh planning
Density
521
TCO criteria
Upfront cost
Recurring cost
Roadmap criteria
Scalability
Legacy compatibility
Scalability
Upgradability
Scalability
Service life
Upgradability
Legacy compatibility
Upgradability
Service life
Service life
Legacy compatibility
Job title
Industry
Expert B
Sr database engineer/information
technolgy manager
Technical marketing engineer
Expert C
Expert D
Expert E
Software engineer
Validation engineer
Platform validation engineer
Expert F
Small organization
Survey respondent
1 small organization
Survey respondent
2 small organization
Large organization
Survey respondent
3 large organization
Survey respondent
4 large organization
Survey respondent
5 large organization
Survey respondent
6 large organization
Survey respondent
7 large organization
Solutions specialist
Expert A
Server administrator
Years of
experience
27
5
7
11
7
5
10
11
Sr database manager
11
Sr database manager
Telecommunications
10
IT manager
Data center managers
Data center managers
Data center managers
10
12
Table AI.
BPMJ
17,3
522
Table AII.
Criteria preferences
RAS
Performance
TCO
Roadmap
Small environment
Large environment
0.22
0.41
0.14
0.22
0.32
0.25
0.25
0.19
Figure A1.
Small business
criteria preferences
Figure A2.
Large business
criteria preferences
Table AIII.
RAS sub-criteria
preferences
Serviceability
Reliability
Manageability
Availability
Small environment
Large environment
0.24
0.26
0.08
0.42
0.18
0.27
0.24
0.31
IT infrastructure
refresh planning
523
Figure A3.
Small business RAS
sub-criteria preferences
Figure A4.
Large business RAS
sub-criteria preferences
Energy efficiency
Transactions/second
Density
Small environment
Large environment
0.35
0.37
0.38
0.28
0.53
0.19
Table AIV.
Performance
sub-criteria preferences
Figure A5.
Small business
performance sub-criteria
preferences
BPMJ
17,3
524
Figure A6.
Large business
performance sub-criteria
preferences
Table AV.
TCO sub-criteria
preferences
Upfront costs
Recurring costs
Small business
Large business
0.35
0.65
0.54
0.46
Figure A7.
Small business roadmap
sub-criteria preferences
Figure A8.
Large business roadmap
sub-criteria preferences
Table AVI.
Roadmap sub-criteria
preferences
Scalability
Legacy compatibility
Upgradeability
Service life
Small business
Large business
0.23
0.20
0.26
0.30
0.26
0.18
0.20
0.36
IT infrastructure
refresh planning
525
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.