Skip to main content

Full text of "A comprehensive methodology for computer-family selection"

See other formats


NPS-5--90-008 


NAVAL POSTGRADUATE SCHOOL 

Monterey, California 



A COMPREHENSIVE METHODOLOGY FOR 
COMPUTER-FAMILY SELECTION 

by 

Moshe Zviran 

// 

March 1990 


Approved for public release; 
Distribution unlimited. 


D ^ na red For: 


FedDocs Postgraduate School 

D 208.1U/2 ^ CA. 93943 

NP.s~s4-QO-008 


Naval Postgraduate School 
Monterey, California 

RADM R. W. West, Jr. Harrison Shull 

Superi uendent Provost 


The research summarized herein was accomplished with funding 
provided by the Research Council of the Naval Postgraduate School. 

Reproduction of all or part of this report is authorized. 
This report was prepared by: 


NAVAL POSTGRADUATE SCHOOL 
MONTEREY CA 93943-5101 


UNCLASSIFIED 


SECURITY CLASSIFICATION OF THIS PAGE 


REPORT DOCUMENTATION PAGE 


Form Approved 
OMB No 0704-0188 


la REPORT SECURITY CLASSIFICATION 

Unclassified 


lb RESTRICTIVE MARKINGS 


2a SECURITY CLASSIFICATION AUTHORITY 


2b DECLASSIFICATION /DOWNGRADING SCHEDULE 


3 distribution /availability of report 

Approved for public release; distribution 

is unlimited 


0. PERFORMING ORGANIZATION REPORT NUMBER(S) 

NPS-54-90-008 


5 MONi T ORlNG ORGANIZATION REPORT NUMBER(S) 


6a NAME OF PERFORMING ORGANIZATION 

Naval Postgraduate School 


6b OFFICE SYMBOL 
(If applicable) 

54 


7a NAME OF MONITORING ORGANIZATION 

Naval Postgraduate School 


6c ADDRESS (City, State, and ZIP Code) 

Monterey, CA 93943-5000 


'b ADDRESS(Gry State, and ZIP Code) 

Monterey, CA 93943-5000 


8a NAME OF FUNDING /SPONSORING 
ORGANIZATION 

NPS Research Foundation 


8b OFFiCE SYMBOL 
(If applicable) 


CUREMENT INSTRUMENT IDENTIFICATION NUMBE : 

O&MN, Direct Funding 


8c. ADDRESS (City, State, and ZIP Code) 

Naval Postgraduate School 
Monterey, CA 93943-5000 


10 


: UND>NG NUMBERS 


PROGRA* 

ELEMEV 


PROJECT 


TASK 

NO 


vVOR^ UNIT 
ACCESSION NO 


1 1 TITLE (Include Security Classification) 

A Comprehensive Methodology for Computer - Family Selection (Unclassified) 


12 PERSONAL AUTHOR(S) 

Moshe Zviran 


13a TYPE OF REPORT 

Technical Report 


^3b TIME COVERED 
FROM TO 


DATE OF REFOPT (Year Month. Day) 

March 1QQQ 


5 p AGE COl 


-LL 


16 SUPPLEMENTARY NOTATlOr 


COSA-, CODES 


: !ElD 


GROUP 


GPOl 


18 SUBJECT TERMS (Continue on reverse if necessary and identify b> twoci number) 

Computer - family, Computer selection, 
Analytic Hierarchy Process 


19 ABSTRACT .Continue on reverse if necessary and identify by block number) 

This paper presents a selection methodology for a computer-family. The 
proposed methodology incorporates the Analytic Hierarchy Process in the evaluation 
procedure and aims at helping organizations in selecting a family of computers from the 
a manufacturer's product line, rather than a specific computer model. 

The practice of computer selection and the existing solutions for a computer- 
family selection procedure are briefly described. Then, Saaty's Analytic Hierarchy 
Process is presented and incorporated into the selection methodology. The result is a 
structured and comprehensive methodology that allows decision makers to rank the 
alternatives more objectively and select a computer-family that best fits the needs of 
the entire organization. Illustrative examples are embedded in the text to demonstrate 
the application of the various steps in the proposed methodology. 


,-. AVAILABILITY OF ABSTRACT 

Cg J'.C.ASS F ED/'JNLiMiTED □ SAVE AS P.-' □ qt r JSEPS 


22a ".-- . RESPONSIBLE INDIVIDUAL 

Moshe Zviran 


: • ABSTRACT SECURI 

Unclassified 


CLASS FlC 


22b Tf ■ Nl (include Area i 

408-646-248^ 


54Zv 


DD Form 1473, JUN 86 


':ons are ob 
U102-LF-0 






A Comprehensive Methodology for 
Computer-Family Selection 


By 


Moshe Zviran 

Department of Administrative Sciences 

Naval Postgraduate School 

Monterey, CA. 93943 

Tel: (408) 646-2489, Bitnet: 5046P(a NAVPGS 


March 1990 


A Comprehensive Methodology for Computer-Family Selection 


Abstract 

This paper presents a selection methodology for a computer-family. The 
proposed methodology incorporates the Analytic Hierarchy Process in the evaluation 
procedure and aims at helping organizations in selecting a family of computers from the 
a manufacturer's product line, rather than a specific computer model. 

The practice of computer selection and the existing solutions for a computer- 
family selection procedure are briefly described. Then, Saaty's Analytic Hierarchy 
Process is presented and incorporated into the selection methodology. The result is a 
structured and comprehensive methodology that allows decision makers to rank the 
alternatives more objectively and select a computer-family that best fits the needs of 
the entire organization. Illustrative examples are embedded in the text to demonstrate 
the application of the various steps in the proposed methodology. 


Keywords : Computer-family, Computer selection, .Analytic Hierarchy Process. 


Acknowledgement : The author wishes to acknowledge the support from the NPS 

Foundation Research Program. 


2 
1. Introduction 

The advent of computers and data communications technology has brought about 
major changes in organizations' computerization process. Motivated by increases in 
their size, sophistication and geographic dispersion, many organizations distribute or 
decentralize their information systems [Ahituv and Neumann, 1986; Ahituv et al. 1989; 
Cash et al, 1988]. This phenomenon requires a close control of the computer 
acquisition process in order to maintain maximum compatibility among dispersed 
systems. Thus, rather than selecting a specific machine for known and identified needs, 
organizations are often faced with the problem of selecting a group of systems that will 
work in harmony, sharing files and data. Such a group of systems is called a computer- 
familv and is defined as: 


Computers of the same type, consisting of several models from the same 
manufacturer's product line, ranging from microcomputer to mainframe, with 
fill compatibility in the operating system and the system 's software, to enable 
transfer of application software from one family member to another without 
change [Borovits and Zviran, 1987]. 


An example of a computer-family is Digital Equipment Corporation's (DEC) 
following line of products: VAX 8978, 8974, 8842. 8840 and 8820 mainframe computers, 
VAX 8810, 6240 and 6230 super-minicomputers. VAX 6220 and 6210 minicomputers 
and VAX 8250, Microvax 2000 and 3600 and Microvax II as super-microcomputers. 
Another example consists of Prime's 6650 and 6350 mainframe computers, Model 6150 
super-minicomputer, Models 4450 and 4150 minicomputers, and Prime models 4050, 
2455. 2450 and 2350 as super-microcomputers. 


3 

Compatibility in hardware and software precludes the system integration problem 
encountered with mismatched systems. The benefits of system-wide compatibility are 
exemplified by the ability to transfer application software from one family member to 
another using a common operating system. 

This paper concerns the problem of computer-family evaluation and selection. It 
describes the existing methods for computer selection, presents Borovits and Zviran's 
(1987) generic methodology for computer-family selection and Saaty's (1977) Analytic 
Hierarchy Process. It, then, proposes a comprehensive methodology to deal with this 
issue. A hierarchy of selection criteria is developed and the application of various steps 
in the proposed methodology is demonstrated throughout the paper. 

2. Computer Selection Procedures 

The traditional computer selection process consists of the following stages: 

• Analyzing the requirements and computing needs of the organization. 

• Determining and defining the requirements for the computer system. 

• Sending the request for proposal (RFP) to qualified vendors. 

• Screening, evaluating, validating and comparing the proposals. 

• Selecting the best alternative. 

[Timmreck, 1973; Joslin, 1977; Borovits, 1984; Borovits and Zviran, 1987; Shoval and 
Lugasi, 1987]. 

The fourth stage, focusing on the actual screening and evaluation of the 
proposals, is the core of the selection procedure. A variety of models and methods for 
this stage have been intensively discussed in the literature. Figure 1 outlines the existing 


4 
computer selection methodologies and the basic reference for each of them (a 
summarized description of these methodologies can be found in: Borovits, 1984; 
Borovits and Zviran, 1987; Shoval and Lugasi, 1987). 


Insert Figure 1 about here 


These methods, however, address the problem of selecting a specific computer or 
computers to meet specific and known requirements. None of them provides a tool to 
evaluate and select a computer-family as defined above. 

3. Computer-Family Selection 

Selecting a computer-family is more complex than selecting a specific computer 
system. In selecting a computer-family, an organization cannot evaluate competing 
products on a one-to-one basis (e.g., DECs VAX 6230 and Prime 6150), but rather 
focus on a comparison of groups of computers with similar characteristics (e.g., DECs 
versus PRLME*s mainframes, super-minicomputers, etc.). 

Borovits and Zviran (1987) have first tackled this issue and proposed a generic 
methodology for the selection of a computer-family. Their methodology consists of the 
following ten steps: 

1. Identification of possible vendors and manufacturers 

2. Preliminary elimination of irrelevant candidates 

3. Determination of mandatory requirements 

4. Examination of vendor's compliance with mandatory requirements 


5 

5. Setting quantitative and qualitative criteria and respective weighting-scales 

6. Writing the RFP to be addressed to selected vendors 

7. Receiving, comparing and analyzing bids 

8. Drawing up a final list of vendors 

9. Benchmarks for performance of hardware and software 

10. Final conclusions and selection of the best computer-family 

This methodology provides a framework for carrying out the computer-family 
selection process. It has, however, two major drawbacks: 

a. It does not encompass an objective weighting technique for setting the weighting 
-scale for the qualitative and quantitative criteria (step 5). It rather 
addresses the need of doing so and proposes a list of relevant selection criteria 
that should be considered. The relative weights for these criteria are assigned 
subjectively. 

b. It suggests the use of the weighted scoring method for comparing and analyzing 
bids (step 7). A major drawback in this method is that it might be influenced by 
subjective considerations. Thus, using subjective weighting and scoring can 
reduce the overall effectiveness of the process. It also does not allow an 
examination of consistency by the evaluators. 

These two issues are addressed in the proposed methodology. An objective 
weighting and scoring technique - Saaty's Analytical Hierarchy Process - is integrated 
into the selection methodology to improve the selection process. Thus, the new 


6 
methodology is more comprehensive and aims at aiding decision makers in the 
computer-family selection process. 

4. The Analytic Hierarchy Process 

The Analytic Hierarchy Process (AHP) was introduced by Saaty (1977) as a 
method for assessing the importance of a large number of interacting factors, develop 
priorities among the factors and choose a best alternative in an objective manner 
[Saaty, 1977, 1981, 1982]. 

The method is based on a pairwise comparison between all relevant factors. 
In each pairwise comparison, a decision maker evaluates two factors and answers the 
question: "Which of the following two factors dominates the other, and by how much ?". 
The first part of the question is clearly an ordinal question, while the second part is a 
cardinal one, requiring a numerical input. The answer is based on a nine-point 
numerical scale, as defined by Saaty (1977) and presented in Figure 2. 


Insert Figure 2 about here 


The answers to these evaluations comprise the input data for a comparison matrix. 
The size of this matrix for n factors is rixn. Each cell represents a pairwise comparison 
between two factors, showing the relative contribution (to the subject of comparison) of 
the /th element as compared to the /th element. The matrix has positive entries 
everywhere and satisfies the reciprocal property, i.e., a M = 1/a,., . Therefore, when the 
//th element of the matrix is specified, the //th position is automatically determined by its 


7 
reciprocal value. Thus, the number of pairwise evaluations required for n factors is 
^(n 2 -n). Figure 3 depicts an example of a comparison matrix of six factors. 


Insert Figure 3 about here 


After a comparison matrix is filled, its eigenvector corresponding to the largest 
eigenvalue is calculated and normalized so that the total sum of its elements is 1. The 
values of this normalized eigenvector (right column in Figure 3) constitute the factors' 
relative weights. 

Another matter of concern is the quality of the answers provided in the 
comparison matrix and, in particular, the problem of consistency. This is assessed by 
considering whether a^ = (a u )*(a kJ ) holds for all triplets. The consistency ratio (CR) is 
calculated for the maximum eigenvalue and is required to be less than 0.1 for acceptable 
consistency. 

Seidmann and Arbel (1985) present an application of the AHP to the process of 
microcomputer selection. They analyze a large number of attributes to compare 
microcomputers from several vendors and provide a case study to demonstrate the 
applicability of their method. Their use of the AHP technique facilitates the 
determination of both weights and scores for each attribute for each alternative, using 
matrices to perform pairwise comparisons between alternatives. The total number of 
matrices in their example equals the number of attributes and the dimension of every 
matrix is the number of alternatives. Once all weights and scores are obtained, the final 
score of each alternative was calculated using the weighted scoring technique. 


8 
5. A Comprehensive Methodology for Computer-family Selection 

The proposed methodology is an elaboration of Borovits and Zviran's 
methodology. It is based on incorporating Saaty's Analytic Hierarchy Process into the 
process of weighting the selection criteria (step 5) and during the evaluation of 
competing computer-families (step 7). Figure 4 presents the proposed methodology 
where the AHP technique is incorporated into steps 5.6 and 7.3. 


Insert Figure 4 about here 


The application of the AHP to steps 5 and 7 of Borovits and Zviran's 
methodology will make the resolution of ranking and weighting alternatives less 
arbitrary. In step 5, the AHP allows a decision maker to objectively create a prioritized 
and weighted list of criteria. At each level of the hierarchy, every criterion can then be 
compared to all the others in its group, on a one-to-one basis. Using the scale and 
descriptions from Figure 2, a score for each pairwise comparison is obtained. These 
scores are inserted into a comparison matrix to compute the relative weight of each 
criterion by Saaty's method, as well as the consistency ratio. 

Step 7 consists of the process of receiving, comparing and analyzing bids. This 
represents a second opportunity for incorporating the AHP technique into the selection 
process. After assigning each relevant model from each proposed computer-family 
to a category' (e.g., mainframe, supermini, mini, micro), each category is evaluated in 
accordance with the criteria established in step 5. The advantage in applying the AHP 
to this step is achieving greater objectivity as categories of computers from different 


9 
manufacturers' product lines are evaluated on a one-to-one basis. 

Following is a step-by-step description of the proposed methodology: 

Step 1 : Identification of possible vendors and manufacturers. This step involves a 
search of all vendors whose product-lines might suit the organization's needs, in 
accordance with the definition of a computer-family. The output of this step is an initial 
list of vendors whose product-line may suit the organization's needs. 

Step 2 : Determination of mandatory requirements . Mandatory requirements 
define the basic features that are required from a computer-family. These requirements 
are derived from the basic definition of a computer-family as well as from the 
organization's information systems (IS) policy. The output of this step is a set of 
requirements (e.g., full compatibility of system's software, ability to upgrade each model 
to a higher one without change in software and operating procedures, etc.), which are 
considered as prerequisites for a vendor's candidacy. 

Step 3 : Examination of vendor's compliance with mandatory requirements. Based 
on the mandatory requirements, information regarding each vendor's compliance with 
these requirements is obtained (e.g., by a questionnaire) from all potential vendors and 
examined by the selection team. 

Step 4 : Preliminary elimination of irrelevant candidates. The list of vendors 
(output of step 1) is screened and those suppliers that do not comply with the 
mandatory requirements are winnowed out. The vendors remaining after this 
elimination procedure constitute the mailing list for the Request For Proposals (RFP). 

Step 5 : Setting quantitative and qualitative criteria and respective weighting scale. 
This step focuses on establishing the evaluation framework, within which all bids will be 


10 
analyzed. In order to select a computer-family that best fulfills its requirements, an 
organization must designate the qualities that will be used to compare the computer- 
families. These qualities, or characteristics, are called selection criteria. 

All criteria used in the evaluation process can be sorted in a hierarchical scheme, 
as illustrated in Figure 5. The top of this hierarchy is denoted as "Total score of a 
computer-family". The second level consists of the division to qualitative and 
quantitative criteria. The next level within the quantitative criteria defines the categories 
of computers and the subsequent levels define specific attributes by which the competing 
families will be evaluated. Criteria at each level are the descriptors of a criterion of the 
next higher level. The lowest level consists of atomistic elements which describe specific 
characteristics and by which the specific computer models are to be evaluated. 


Insert Figure 5 about here 


Step 5 is broken into the following seven sub-steps: 

Step 5.1. Prioritize the overall importance of qualitative versus quantitative criteria. 
The criteria used to evaluate computer-families are either quantifiable or non- 
quantifiable. The qualities that are not quantifiable are referred to as qualitative 
criteria while those characteristics that are quantifiable and measurable by an 
established standard are called quantitative criteria. Because both qualitative and 
quantitative criteria used in the evaluation process, the first step is the determination of 
the relative weights, or percentage of the total score, for each of these groups of criteria. 
This is a subjective decision, and since only two factors are involved (qualitative criteria 
and quantitative criteria), it is made without the use of the AHP. 


11 

Step 5.2. Set qualitative criteria. Qualitative criteria are used to describe general 
characteristics of a vendor or a computer-family which, although nonquantifiable, are 
important to the overall evaluation process. Figure 6 illustrates a multi-level hierarchy 
of qualitative criteria. 


Insert Figure 6 about here 


Step 5.3. Select applicable computer categories. In accordance with the definition 
of a computer-family, vendors are expected to propose a wide variety of elements rather 
than a single computer. This raises a problem of comparing proposals from different 
vendors. To overcome this problem, a scheme of computer categories is to be 
established, each of which represents differences in computing power and major 
hardware characteristics. This will enable a classification of each proposed system into a 
specific category and evaluation of categories of computers rather than specific systems. 

Thus, the development of the quantitative branch of the selection criteria scheme 
starts with the definition of applicable computer categories. An example of such a 
classification might consist of: 

- Mainframe 

- Super-minicomputer 

- Minicomputer 

- Super-microcomputer 

- Microcomputer 

Step 5.4. Set quantitative selection criteria. Quantitative selection criteria describe 
the major measurable characteristics for each computer and are applied to each of the 


12 
categories designated in Step 5.4. As proposed by Borovits and Zviran (pp. 110-111), 
the quantitative criteria should usually address the following common issues: hardware, 
software, communication, conversion and environment. These criteria constitute the 
fourth level in the hierarchical scheme. 

Step 5.5. Select sub-criteria for each criterion, down to the lowest level. 
Elaborate on each criterion, set in the previous step, and develop appropriate sub- 
criteria. The sub-criteria selected represent a break-down of each criteria and should be 
valid and meaningful items of comparison that can be applied to the actual evaluation of 
the proposed computer-families. An example of a criteria pertaining to hardware 
characteristics is presented in Figure 7. 


Insert Figure 7 about here 


Step 5.6. Prioritize and weight all categories, criteria and sub-criteria. 
This step focuses on assigning a relative importance to each category, criterion and sub- 
criterion, using Saaty's AHP process. As already described, the AHP method consists of 
a pairwise comparison of elements at each level of the hierarchy. Every element being 
compared is rated against all other elements in the same level, on a one-to-one basis. A 
value is obtained, based on the scheme presented in Figure 2, and inserted into a 
comparison matrix. The size of the matrix equals the number of elements being 
compared and recommended to be limited to 5-9 items (Saaty, 1982). The normalized 
eigenvector of matrix generates a relative weight for each element. The total value of all 
the weights generated for each group being compared is 1. 


13 

As an example, assume that the classification of computers (step 5.3) yielded 
three categories - mainframe, minicomputers and microcomputers - which are to be 
prioritized. To determine the relative weight for each category, a decision maker has to 
compare each category with the other two, one at a time, and assign a numerical value 
that best represents the intensity of importance of one category over the other. 

The numerical values are inserted into a comparison matrix and their reciprocals 
are calculated for the corresponding cells, as illustrated in Figure 8. Then, the 
normalized eigenvector is calculated to represent the relative importance of each of the 
items being evaluated, the consistency ratio is computed to ensure the consistency of all 
responses and weights. 


Insert Figure 8 about here 


Step 5.7. Calculate the absolute weights for all criteria and sub-criteria. The 
absolute weight for a criterion is computed by multiplying its relative weight by the 
relative weight of each of its predecessors in the hierarchy, or by the absolute weight of 
its immediate predecessor. 

Figure 9 presents an example of absolute weight calculation. 


Insert Figure 9 about here 


The process demonstrated in Figure 9 is completed for each criterion in the 
hierarchy of a computer-family. These weights represent the maximum absolute values 


14 
that a given computer-family can score in the evaluation process (step 7). Absolute 
weights of sub-criteria at the lowest level of the hierarchy are used in calculating the 
absolute scores in step 7. 

Step 6 : Writing the RFP to be addressed to selected vendors. The RFP consists of 
a summary list of specific requirements according to which vendors will be asked to 
write their proposals. Following the evaluation scheme, the RFP should include two 
parts. The first focusses on quantitative criteria and relates to each computer model 
within a proposed family. The second is more general in nature and addresses 
qualitative criteria. It concentrates upon issues such as uniformity and transferability of 
systems software, conversion of present applications software to the new computer- 
family, environmental considerations, etc. The RFP is mailed to vendors according to 
the mailing list created as the output of step 4. It is required that the bids be submitted 
in writing and it is expected that they will conform to the style indicated in the RFP, so 
that the selection process will not be affected by style of expression and use of selling 
techniques. 

Step 7 : Receiving. Comparing and analyzing bids. In response to the RFPs, bids 
for proposed computer-families will have been received. These bids have to be analyzed 
and evaluated as a basis for selecting the highest rated computer-families for final 
evaluation in steps 8, 9, and 10. This evaluation is performed in six sub-steps: 

Step 7.1. Assign each relevant model from each proposed computer-family to a 
category, according to predetermined criteria. Computer categories have been established 
in step 5.3. This stage focuses on classifying each model from each proposed computer- 
families to an appropriate computer category so it is evaluated by the criteria already set 
for that category. 


15 

The determination as to which category a computer will be placed is based on, 
but not limited to, such factors as CPU performance, memory size, external storage 
capacity, number of disk drives, cost, etc. The decision as to what factors will constitute 
placement into a particular category will have been determined when the categories 
were selected in Step 5.3. 

Step 7.2. Design comparison tables for each category. Once all proposed models 
have been classified to categories, comparison tables are designed. These tables 
summarize the characteristics of each proposed model within each given category. A 
separate table is designed for each category. An outline is presented in Figure 10. 


Insert Figure 10 about here 


Step 7.3. Evaluate each computer model in accordance with the criteria established 
in Step 5. This stage provides a second opportunity for incorporating the AHP 
technique in the selection procedure. During this stage, each computer model is 
evaluated within the category he is assigned to. If a vendor proposes more than one 
model in a given category, all proposed models are evaluated. Computer models in a 
given category are compared by criterion, using the AHP technique. A pairwise 
comparison of these models is performed and a value, based on the AHP technique, is 
obtained. After all values for a criterion are obtained, the relative scores and 
consistency ratio are calculated using the procedure described in step 5.6. All scores are 
recorded in an evaluation table as outlined in Figure 10. 


16 
Step 7.4. Calculate the absolute score for each criterion and each computer model. 
Based on the evaluation tables designed in the previous step, the absolute score is 
calculated. It is computed by multiplying the relative score (outcomes of step 7.3) and 
the absolute weight of each criterion, as calculated in step 5.7. A formal representation 
of this computation is: 

s, = vw„ 

Where: 

Sj = absolute score attained by a specific computer model, for 

a given criterion j 
R„ = relative score attained by a computer model in category i, 

for criterion j, on scale of 0-1 
W,j = absolute weight of criterion j in category i (as calculated 
in Step 5.7), on scale of 0-1 
In each category, S, is calculated for all computer models. Based on these scores, 
a comparative table is drawn up, showing the absolute scores attained for each criteria 
by each computer model. This table uses the same outline as illustrated in Figure 10. 

Step 7.5. Calculate the total score for each computer model. Based on the 
comparative tables (output of previous step), the total score for each computer model in 
each category is computed over all criteria as: 

j 

Where: 

S = overall score attained by a specific computer model 
S = absolute score attained b\ a specific computer model 
(output of step 7.4), for a given criterion j. 


17 

Step 7.6. Calculate the total score for each computer- family. The final stage in 

evaluating the bids focuses on scoring the computer-families. The overall score for a 

computer-family consists of a summation of the highest absolute score attained by a 

member of a family member in each category. Based on the comparative tables (output 

of the previous step), the total score for each computer-family is computed as: 

T = 2 S k 
k 

where: 

T = total score for a computer-family 

S k = best absolute score attained by a specific family member, 

in category k. 

Step 8 . Drawing up a final list of vendors. On the basis of the final scores attained 
by each computer-family, the selection committee is able to disqualify irrelevant 
computer-families, and select up to three or four vendors most likely to succeed. These 
computer-families are then further tested to ensure they have the proper capabilities and 
characteristics. 

Step 9. Benchmarks for performance of hardware and software. A benchmark, in 
the context of this discussion, is a set of live tests designed to examine the characteristics 
and actual performance of the proposed systems (hardware and software). One category' 
of benchmark tests aims at verifying cardinal characteristics of the proposed computer 
families (e.g., uniformity of the operating system and application software, ease of 
converting existing applications to the proposed family). Another type of benchmark 
tests refers to examining the systems performance using common production measures. 
Examples of such measures include total throughput and transaction volume load, which 


18 
delineate the expected capacity of the system to handle the anticipated average 
workload. Another measure, peak load handling, refers to the system's response to 
temporary added load. The selection of issues and criteria to be tested is performed 
according to their importance for an organization using the relative weights already 
assigned. 

Step 10 . Final conclusions and selection of the best computer-family. After 
benchmarks have been performed and all essential characteristics of a proposed 
computer-family have been deemed satisfactory, a selection committee will review and 
reconsider the relevant scores assigned to each competing computer-family. 

Finally, the committee will pick the best as the one recommended to be an 
organization's computer-family. The recommendations will then be submitted to an 
organization's management for approval and adoption. A problem faced by those 
involved in the selection process is how to compare criteria and how to prioritize them 
according to their importance to the decision making process. There will be a large 
number of criteria, some quantifiable and others non-quantifiable, whose importance to 
the selection process will be compared with each other. 

6. Discussion and Conclusion 

The need to develop a comprehensive methodology for computer-family selection 
arises from the trend towards distributing computing resources. Organizations with 
distributed or decentralized systems, or in the process of carrying out decentralization of 
computing resources, should be in a position to evaluate and select a computer-family 
rather than a specific computer model. 


19 

Selecting a computer-family will ensure a uniform computing environment for the 
entire organization. This environment provides full compatibility in both hardware and 
systems software and minimizes the cost of systems integration. Moreover, the ability to 
transfer application software from one family member to another without any change 
avoids duplication in software development and lays the foundation for coordinated 
development and implementation of consistent and organization-wide information 
systems. Another advantage lies in the ability to focus on a one-time effort for the 
evaluation and selection process. 

The process of selecting a computer-family is a complex procedure. The goal for 
a decision maker, responsible for selecting a computer-family, is to select the correct 
line of products for an organization rather than a specific computer. Because of the 
complexity of the selection process, a formalized methodology makes the process more 
structured and objective. 

The methodology presented here provides a comprehensive framework to carry 
out the selection process. It allows the designation of a hierarchy of selection criteria, 
based on the organizational needs. Once criteria have been selected, they are 
objectively prioritized and weighted using the AHP technique, establishing their net 
value and absolute weight for the overall evaluation process. Each computer model 
within proposed computer-families is then evaluated and scored separately, in 
accordance with the prioritized and weighted criteria. The total score for a computer- 
family is based on the aggregation of final, absolute, scores of the best performing family 
members in each category. 


20 
By following the procedure presented in this paper, the process of selecting a 
computer-family is made reliable and objective. The end product of this process, a 
computer-family that best meets the needs of an organization, may be chosen with the 
knowledge that the correct computer-family was selected. 


21 
References 

Ahituv N. and Neumann S., Principles of Information Systems Management, 

Wm. C. Brown, Dubuque, Iowa, Third Edition, 1990. 
Borovits I., Management of Computer Operations, Prentice-Hall, Englewood Cliffs, 

NJ, 1984. 
Borovits I. and Zviran M, "Computer-Family Selection for Organizational Systems", 

Information and Management, Vol. 12, No.3, (March 1987), pp. 107-115. 
Ein-Dor P., "A Dynamic Approach to Selecting Computers", Datamation, Vol. 23, 

No. 6, (June 1977), pp. 103-108. 
Joslin E.O., Computer Selection, The Technology Press, Arlington, VA, 1977. 
Roenfelt R.L. and Fleck R.A., "How Much Does a Computer Really Cost ?", Computer 

Decisiom, Vol. 7, No. 5, (November 1976), pp. 75-79. 
Saaty T.L., "A Scaling Method for Priorities in Hierarchical Structures", Journal of 

Mathematical Psychology*, Vol. 15, No. 3, (1977), pp. 234-281. 
Saaty T.L., The Analytic Hierarchy Process, McGraw-Hill, New York, NY, 1981. 
Saaty T.L., The Logic of Priorities, Kluwer-Nijhoff, Boston, MA, 1982 
Seidmann A. and Arbel, A., "Microcomputer Selection Process for Organizational 

Information Management", Infomiation and Management, Vol. 7, No. 5, 

(December 1984), pp. 317-329. 
Sharpe W.F., The Economics of Computers, Columbia University Press, New York, 

NY, 1969. 
Shoval P. and Lugasi Y., "Models for Computer System Evaluation and Selection", 

Infomiation and Management, Vol. 12, No. 5, (March 1987), pp. 117-129. 
Shoval P. and Lugasi Y., "Computer Systems Selection: The Graphical Cost-Benefit 

Approach", Information and Management, Vol. 15, No. 3, (October 1988), 

pp. 163-172. 
Timmreck E. M., "Computer Selection Methodology", Computing Surveys, Vol. 5, No. 4. 

(December 1973). pp. 199-222. 


Selection Method 

Basic Reference 

Weighted Scoring 

Sharpe, 1969 

Cost-Value 

Timmreck, 1973 

Dynamic Approach 

Ein-Dor, 1977 

Present Value 

Roenfelt and Fleck. 1976 

Cost-Effectiveness 
Ratio 

Joslin, 1977 

Requirement Costing 
Technique 

Borovits, 1984 

Eigenvector Model 

Seidmann and Arbel, 1984 

Lexicographical 
Ordering 

Ahituv and Neumann, 1986 

Multi-Attribute 
Utility Model 

Shoval and Lugasi, 1987 

Efficient-Frontier 
Model 

Shoval and Lugasi, 1988 


Figure 1: Existing methods for computer selection 



Level 

of 




Impon 

ance 

Definition 


Explanation 

1 


Equal importance of 
the two factors 


The two factors 
contribute equally 

3 


Weak importance of 
factor i over factor j 


Experience and judgment 
slightly favor one factor 
over another 

5 


Strong importance of 
factor i over factor j 


Experience and judgment 
strongly favor one factor 
over another 

7 


Very strong importance of 
factor i over factor j 


One factor is strongly 
favored and its dominance 
is demonstrated in practice 

9 


Absolute importance of 
factor i over factor j 


The evidence favoring one 
factor over another is of 
the highest possible order 
of affirmation 

2,4,6.8 


Intermediate values between 
two adjacent scale values 


Compromise is needed 
between two levels 

Reciprocals 

If factor i has one of the pre 

ceding 

numbers assigned to it when 



compared with factor j, then 

factor 

1 is assigned with the reciprocal 



value when compared with factor i 



Figure 2: The comparison scale for Saaty's method 





Factors 




Eigen- 


A 

B C 

D 

E 

F 

vector 

Factor A 

1 

2 3 

2 

1/2 

1/3 

.15 

Factor B 

1/2 

1 2 

2 

1/2 

1/3 

.12 

Factor C 

1/3 

1/2 1 

1/2 

1/4 

1/6 

.05 

Factor D 

1/2 

1/2 2 

1 

1/3 

1/4 

.08 

Factor E 

2 

2 4 

3 

1 

1/2 

.23 

Factor F 

3 

3 6 

4 

: 

1 

.37 

1.00 

Maximum e 

igenval 

ue for this matrix = 

6.10 



Consistency 

Ratio 

= .016 






Figure 3: Scoring factors using Saaty's AHP: an example 


Step 1. Identification of possible vendors and manufacturers. 

Step 2 Determination of mandatory requirements. 

Step 3 Examination of vendors' compliance with mandatory requirements. 

Step 4 Primary elimination of irrelevant candidates. 

Step 5. Setting quantitative and qualitative criteria and respective scales. 


5.1. 

Prioritize overall importance of qualitative 


and quantitative criteria. 

5.2. 

Set qualitative criteria 

5.3. 

Select applicable computer categories. 

5.4 

Set quantitative criteria 

5.5. 

Select sub-criteria for each criterion down 


to the lowest level. 

5.6. 

Prioritize and weight all categories, 


criteria and sub-criteria. 

5.7. 

Calculate the absolute weights for all 


criteria and sub-criteria. 


Step 6. Writing the RFP to be addressed to selected vendors. 
Step 7. Receiving, comparing, and analyzing bids. 


7.1. Assign each relevant model of computer from a 
proposed computer-family to a category. 

12. Design comparison tables for each category 

73. Evaluate each computer model in accordance 
with criteria established in Step 5. 

7.4. Calculate the absolute score for each criterion 
and each computer model 

7.5. Calculate the total score for each computer model 

7.6. Calculate the total score for each computer family 


Step 8. Drawing up a final list of vendors. 

Step 9. Performance of hardware and software benchmarks. 

Step 10. Drawing final conclusions and selection of best computer-family. 


Figure 4: A comprehensive computer family selection methodology: 

A workflow diagram 


LEVEL 1 


LEVEL 2 


LEVTL 3 




LEVTL 4 

LEVTL 5 


LEVEL 6 


LEVEL 7 



|k* inf ramc 
™| Compute r 




™yf 


1 1 

Capacity | 





Disk.5 1 

1~H 


Expandiblll 

:y| 















— L^U 


Speed 



1 

i 

(Mini 

1 Compute r 




Diskette 

Ou»r,..;auve , 


i 1 


readers 







1 Hard-a re 


- 






Te rmlni 1 






Printers 

! 


lKicro 


Ccxrput. c r 

Bockup 1 

1 1 








1 




Vender 
support 

- 






1 
Vendor 

1 Sclt-are 







1 Cornnun l CAt l or. ■ 


KanuJ acturer 
rcput si i or, 

"I i 







:ji.;i*-,ivf 1 



\ Conversion 





|»un».r 
users o: 



compute r- 
t arr^ . ) 

"1 I nv i rorur>-nt. 






So£t««re 

con-.pa file s 
speciali : ing 
in this 
comput t:- 
l ami 1 y 

















Quantitative criteria in levels u-7 are replicated for all computer categories 
These criteria are further broken down (see Figure 7 fc more details) 


Ficure 5: Hierarchy of criteria 


Level 1 


Level 2 


Level 3 


Level 4 


Total score 


Qualitative criteria 


Vendor support 


Implementation assistance 
Technical trouble-shooting 
Training 
Documentation 


Vendor reputation 


Spread of use 


User opinions 

Trade journal evaluations 


Number of organizations 
Number of installed systems 


Software houses specializing 
in this computer-family 


Figure 6: Detailed List of Qualitative Criteria 


Level 4 


Level 5 


Level 6 


Level 7 


Hardware 


Memory 


Disks 


Tapes 


Diskette drives 


Data channels 


Terminals 


Printers 


Main 


RAM 


Cache 


Word size 
Standard capacity 
Maximum capacity 
Units of expansion 
Speed 

Capacity 
Expendability 

Capacity 
Expendability 


Maximum number of drives 
Minimum capacity 
Maximum capacity 
Average access time 
Data transfer rate 

Maximum number of drives 
Density 

Read/Write speed 
Data transfer rate 

Maximum number of drives 
Diskette size 
Density 
Read/Write speed 

Minimum number of channels 
Maximum number of channels 
Average transfer rate 


Monitors 


Keyboards 


Dot matrix 


Letter quality 


Line 


Laser 


Monochrome 

Color 

Graphics capabilities 

Number of keys 
Desiim 


speed 
line size 

speed 
line size- 
speed 
line size 

speed 


Figure 7: Detailed List of Criteria, Hardware 



Main- 
-frame 

Mini 
computer 

Micro 
computer 

Relative 
weights 

Mainframe 1 
Minicomputer 1/3 
Microcomputer 1/3 

3 

1 
1/2 

3 

2 
1 

.59 
.25 
.16 

1.00 

Maximum eigenvalue for this 
Consistency Ratio = .043 

matrix = 3.05 




Figure 8: Calculating relative weights for computer categories 


Computer-family total weight = 1.00 
Quantitative criteria relative weight = 0.80 
Mainframe computer relative weight = 0.59 


Absolute weight for mainframe computers: 
(1.00) * (0.80) * (0.59) = 0.472 


Figure 9: Computing the absolute weight of mainframes 


Mainframe comparison table 


Criteria 

I 
Vendor A Vendor B 

i 

l 
Vendor C 

Model A1 

Medel A2 

Model A3 

Model B1 

Model B2 

Model C1 

Crit. 1 
Crit. 2 
Crit. 3 








Figure 10: Outline of a comparative table. 


Distribution List 


Agency No. of copies 

Defense Technical Information Center 2 

Cameron Station 
Alexandria, VA 22314 

Dudley Knox Library, Code 0142 2 

Naval Postgraduate School 
Monterey, CA 93943 

Office of Research Administration 1 

Code 012A 

Naval Postgraduate School 

Monterey, CA 93943 

Library, Center for Naval Analyses 1 

4401 Ford Avenue 
Alexandria, VA 22302-0268 

Department of Administrative Sciences 1 

Code AS 

Naval Postgraduate School 

Monterey, CA 93943 

Professor Moshe Zviran, Code AS/Zv 10 

Department of Administrative Sciences 
Naval Postgraduate School 
Monterey, CA 93943