John F. (John Fralick) Rockart.

Centralization vs decentralization of information systems : an annotated bibliography online

. (page 1 of 2)
Online LibraryJohn F. (John Fralick) RockartCentralization vs decentralization of information systems : an annotated bibliography → online text (page 1 of 2)
Font size
QR-code for this ebook



















- 1\ij^



Center for Information Systenns Research

Massachusetts Institute of Technology

Alfred P Sloan School of l\^anagement

50 Mennonal Drive

Cannbridge. Massachusetts, 02139

617 253-1000


, V by

John Fralick Rockart

Joav Steve Leventer

SLOAN WP 844-76

April 1976



MAY 21 19761


MAY 2 4 1976 i


The field of Distributed Processing is eliciting a great deal
of interest at the current time. The first task undertaken by the
CISR research project on Distributed Processing was one of searching
and culling the literature. An annotated bibliography, representing
the results of this search, is presented below.

Four classes of articles are surveyed, under the headings:

Centralization vs. Decentralization: Issue Discussions
Distributed Systems and Computer Networks
Distributed Data Bases
Organizational Issues

Articles which we found to be of most interest (for our
specific purposes) are marked by an *.


I. Centralization vs. Decentralization: Issue Discussions

Albrecht, : Organization and Management of Information Processing
Leon K. Systems , McMillan, 1973.

This book surveys the general area of information
systems management and organization, pp. 115-120 deal
specifically with issues of Centralization-Decentralization,
stressing in particular organizational and human

Burnett, : "Computer Options: Large Centralized Computer vs.
Dr. G.J. Minicomputers" Unpublished manuscript. Index Systems,
Inc., 1972.

A discussion of the use of minis as stand alone machines
and as parts of larger systems. Provides also a
comparison of large machines and minis along the
dimensions of hardware, software, costs, reliability,
and manufacturers support.

Burnett, : "At Last, Major Roles for Minicomputers" Harvard
Gerald J. Business Review , May-June 1975.

Nolan, Surveys briefly reasons and disadvantages of central i-
Richard L. zation and presents a good table of technical comparison
of computers (large, medium, and mini) along hardware
and software dimensions. Describes four configurations
of computers, ranging from no use of minis to use of
minis exclusively.


Canning, : "Are We Doing the Right Things?" EDP Analyzer , Vol. 13,
Richard G. No, 5, May 1975.

In this, the first in a series of four reports dealing
with the economics of data processing. Canning discusses
the choice and prioritization of D.P. projects. Three
forms of organization for this decision are discussed
(along with examples): Division-Level decision and
Division-Level implementation; Division-Level decision
and Corporate-Level implementation; and Corporate-Level
decision and implementation. The article touches on
some of the major issues in the management of information

* Articles which we found to be of most interest (for our specific
purposes) are marked by an *.


Erikson, : "Determining the Optimal Degree of Computer System

Warren J. Centralization" Working Paper No. 33-1, School of

Business Administration, University of Souther
California, L.A. , 1973.

^^. attempt to determine mathematically, based on queueing
theory, "Whether a multiple computer installation is
cost-effective relative to a single computer installation'
The model developed in the paper relates exclusively
to hardware issues, but includes the issue of communi-
cations costs between separate physical locations.
Includes surveys to determine ranges of values for model

Ware A.

'A Case for Centralized
January 1972.

EDP" Business Automation,

-xplains the trend towards centralization in terms of
lesire for standardization, growth of application
complexity, and shortage of EDP personnel. Reviews
current centralization practices (trends), and suggests
a specific organization of the centralized EDP "which
resembles in miniature, the structure of the environ-
ment which it is established to serve". The functions
v/ithin the "management services" department are
'.uggested as: -business systems application, systems
projects, EDP operations, training and standards,
administrative services.

* Glaser,

'The Centralization vs. Decentralization Issue:
Arguments, Alternatives and Guidelines: Data Base
1970, pp. 1-7.


centralization vs. Decentral
and authorities is discussed
zation include economies of
and EDP dept., availability
Arguments for decentral izati
problems, response to needs.
Hardware and personnel devel
tralization are mentioned,
determining policy. Guideli
to central and division task
underlying principle of loca

ization of staff, equipment
. Arguments for central i-
scale, better control on Co.
of qualified personnel,
on: familiarity with local

local responsibility,
opments which favour decen-
States seven criteria for
nes for allocating tasks
s are provided, with the
1 responsibility.


"Grosch 's Law Revisited'
p. 9.

Computerworld , April 15, 1975,

Grosch 's rebuttal to those who argue that minicomputers
have made his now-famous law obsolete. He re-states
the law in its original form ("Economy is as the square
root of the speed: if you want to do it twice as

cheaply you have to do it four times as fast"), and
suggests that the law proved correct because of the
nature of the behavior of professional users, as well
as its use in pricing. Grosch concludes that: "As
long as we have greedy salesmen and ... programming,
most of the power of even the cleverest machines will
be wasted".




"Speaking of Minis" Datamation , July 1973, pp. 57-59.

Following a brief description of minis, the authors argue
strongly that "An idle computer is not necessarily a
bad thing" and try to show that the use of minis may well
be more useful and less expensive than using a time-
sharing service.

jmfrey, : "The Organizational Side of Centralized EDP Services"
Robert D. Management Control , 21:11-3, January 1974.

Concentrates on organizational issues of centralization.
Includes advantages of centralization, and some basic
difficulties. Argues for centralization, but suggests
some precautions.

* Lassoff, : "Management Considerations: Centralization vs. Decentral-
D.A. ization of EDP Resources" Guide 38 , v. 2, 1974,
pp. 472-487.

Presents a study of a group of executives, "geared to
produce points for consideration" on the subject. Four
environments compared: all combinations of centralization
-decentralization of computer resources, and centrali-
zation-decentralization of development personnel. No
conclusion is offered.

Lowe, : "The Corporate Data Center: Getting It All Together"
Ronald L. Computer Decisions , May 1973.

Presents the steps and alternatives towards "consoli-
dating" the company's EDP activities. A seven step
consolidation plan is presented. Main advantages and
dangers of centralization are discussed.

* Norton, : "Information System Centralization: The Issues"
David P. Harvard Business School Paper, 9-172-286, 1972.

Summarizes arguments made mostly by other authors.

Paper organized by: system operation, system development,

system management.

Streeter, : "Centralization or Dispersion of Computing Facilities:
D.N. IBM Systems Journal , No. 3, 1973.

Attempts to quantify economic and some behavioral aspects
of centralization. Proposes a cost model which includes
reliability considerations. Employs queueing theory
to look into the service-quality questions. Recognizes
behavioral and other subjective issues which draw towards
decentralization, suggesting measuremen:s of "cost of
compartmentalization". Uses many simplifying assumptions.

* Withington, ;
Frederic G.

The Organization of the Data Processing Function ,
Section 7: Multicenter Networks, Wiley Business Data
Processing Library, 1972, pp. 69-79.

An overview of the issues and possible solutions. Details
four forces causing centralization (cost of duplicate
system development, desirability of standard equipment,
desire for uniform management reporting, shortage of
expert personnel) and organizational issues leading to
decentralization. Surveys common alternatives and
compromises, including a "satellite system". The influence
of the parent organization's authority structure and
the effects on the parent organization are discussed.

Withington, :
Frederic G.

"Crystal Balling: Trends in EDP Management",
Infosystems . Volume 20, January 1973.

A shortened, somewhat watered-down version of the same
author's "The Organization of the Data Processing
Function" - Section 7.

Wofsey, : "Centralization vs. Decentralization" in Wofsey:
Marvin M. Management of ADP Systems , Chapter 2, Auerbach 1973,
pp. 25-38.

Another discussion of the argument pro and con central-
ization. Leans towards centralization for economic
reasons. Short discussion of the military command
and control system.


II. Distributed Systems and Computer Networks

Stanford R.

"Distributed Intelligence in Data Communications
Networks", Computer , November-December 1971, pp. 27-32.

An analysis of the ways in which the addition of a
minicomputer can reduce the cost of a computer/communica-
tion network. The author suggests two places where
such a mini could be positioned: at the central site,
taking load off the larger computer, and at a remote site,
as a message concentrator. Advantages obtainable from
both configurations are discussed.

Canning, : "The Emerging Computer Networks", EDP Analyzer ,
Richard G. January 1973,

Discusses the importance to the whole D.P. field of
computer networks such as the ARPANET and TYMNET.

Richard G.

* Canning,
Richard G.

"Distributed Intelligence in
EDP Analyzer , February 1973.

Data Communication'

Discusses the use of minicomputers as front end processors
in data communication. Three case studies presented.
Some arguments against distributed systems rebutted.

"In Your Future: Distributed Systems?", EDP Analyzer ,
August 1973, Vol. 11, No. 8.

A good and thorough article, discussing distributed
processing, its advantages, disadvantages, problems, and
current state. Four major components of distributed
systems identified as: distributed processing, distributed
communications, distributed data bases, and system wide
rules. Predicts that distributed systems will be the
rule by 1980.

* Canning,
Richard G.

"Structure for Future Systems"
Vol. 12, No. 8.

EDP Analyzer , August 1974,

Discussion of three alternative structures: centralized,
hierarchical distributed and network distributed.
Technological trends and their effects are analyzed.
Recommends preparing for the new structures now. Predicts
IBM's support of hierarchical networks. Two case studies

• n :


Chen, : "Distributed Intelligence for User Oriented Computing"
Tien Chi AFIPS . Vol. 41, Part II, Fall 1972, pp. 1049-1056.

A highly technical paper, arguing that new LSI tech-
nology facilitates distributing logic -memory systems,
creating "loosely-coupled polycentric computing
systems". Such systems would be more efficient and
more user oriented, allowing, in particular, efficient
interpretive processing.

Coleman, : "ACCNET - A Corporate Computer Network", AFIPS, Vol 42,
Michael L. 1973, pp. '33-130.

A description of a specific networking of a DEC 10 with
an IBM 370/155. Includes a short description of
other implemented networks, advantages of networks,
and a 56 item bibliography.

David, : "Role of the Minicomputer", Data Management, February
Jon 1975, pp. 16-19.

Surveys the development of minis, discusses distributed
data processing systems (several examples) and suggests
guidelines for getting a minicomputer.

Edited : Datamation , February 1975, pp. 40-56.

This issue includes four articles describing four
different network configurations. The articles are
technical, but give a good overview of the range of
possibilities available with present "state of the
art" technology.

* Edited : "Distributed Computing: A Growing Concept", Infosystems,
August 1975. ^

A survey of the points of view of several hardware
manufacturers associated with the distributed processing
idea (DEC, H-P, etc.). The article is appended by
an excerpt from an address "A Business Approach to
Improving Productivity in the Service Sector", by
R.B. White, Executive V.P., First National City Bank,
New York.


Farber, : "A Sjrvey of Computer Networks", Datamation 18, 4
D-J- (April 1972), p. 36-39.

Give; a good overview of types of networks existing
toda/, their features and problems.

David J.

"Software Considerations
Computer , March 1974, pp.
Arrowhead Workshop).

in Distributed
31-35. (From


1973 Lake

Surveys briefly the issues of interprocess communication,
protection, resource management, and distributed file
systems. Gives examples of systems operating with
the above characteristics.

* Hobbs,

"The Rationale for Smart Terminals", Computer ,
November-December 1971, pp. 33-35.

A short paper arguing for moving processing power
closer to the user (while retaining the data-base
centrally). Suggests guidelines for the degree of
decentralization of "intelligence".

* Joseph,
Earl C.

"Innovations in Heterogeneous and Homogeneous Distri-
buted-Function Architectures", Computer , March 1974,
pp. 17-24. (From Lake Arrowhead Workshop).

Gives a technical survey of types of networks and
distributed architecture machines, their characteristics,
advantages, etc. Includes summary outlines of the

Kneppelt, : "A Simple Distributed Approach to Manufacturing
Leland R. Information Systems", AFIPS , Vol. 43, 1974, pp. 485-490.

Describes an actual implementation of a distributed
system. On-line applications - current status and
control information systems - were implemented on a
minicomputer. Large batch applications, such as long
range planning and manufacturing standards maintenance,
were implemented on a maxi . Files were distributed
between the two systems, with some duplicate information,
and some duplicate files.

* Markowitz,

"A Large Computer Looks at the Distributed Computing
Problems", Behavior Research Methods and Instrumentation ,
Volume 6, No. 2, 1974, pp. 237-240.

Suggests that the facilities needed to complement small
laboratory minicomputers could be supplied remotely
by a large machine via telecommunication. Proposes
three major areas where the large computer can give
such service: remote hardware peripherals, applications
software (e.g. complex data management and data analysis
programs), and programming aids for the small machine.

Describes partial implementation of the "Consistent
System" at MIT.

Marshall, : "Distributed Processing on Wall Street, Datamation ,
Joseph C. July 1973, pp. 44-46.

Describes a mini-network for on-line inquiry and
updating implemented by Bunker-Ramo Corp. The
network, called Telequote III, is used for informing
stock brokers around the country of current stock
process. The newer version of this system. Market
Decision System 7 (MDS-7) is also described.

Peck, : "Effective Corporate Networking, Organization, and
Paul L. Standardization", AFIPS , Vol. 39, Fall 1971,
pp. 561-569.

Deals with networking of multiple ADP centers within
one corporation. Mentions some advantages, and proposes
principles for management and setting standards for
corporate ADP networks.

Riley, : "Minicomputer r.etworks - A Challenge to Maxicomputers?"
Wallace B. in A Practical Guide to Minicomputer Applications ,

Fred F. Coury, pp. 103-109, reprinted from Electronics ,
March 28, 1971.

This article, written by the computer editor of Electronics,
surveys mini-networks and expert opinions on their
advantages and uses. The ability to tailor the computer
network to the particular job, along with cost-effective-
ness, are mentioned as the major advantages. On the
other hand, some experts mention the mini's inferiority
in large calculations and data base handling.

Robbins, : "Distributed Data Processing", Data Management ,
Clark September 1975.

A somewhat disorganized discussion, arguing that the
time of distributed processing has come. Robbins
discusses four levels in the distributed network:
a large data utility, a satellite computer, the
intelligent terminal and the terminal, claiming that
"the processing power at these levels is and will
continue to increase in a dramatic fashion". The
article's major claim is that total cost per terminal/
month is a good measure of the efficiency of computer
systems, and that distributed systems are superior in
this measure to time-sharing systems.



"Computer Networks for Retail Stores", Computer ,
April 1973

A case study of a mini network implemented in a group
of supermarket chains. Emphasis on the in-stores
applications (P.O.S., etc.), and a separate communica-
tions-oriented system for ordering and other functions
on a chain-wide basis.


"Monitoring/Control by Distributing Computing",
Datamation , July 1973, pp. 47-49.

A case study of a network of 45 computers controlling
a Canadian/U.S. oil pipeline that transports over one
million barrels a day. The network, composed of one
host (POP 10) and 44 control minicomputers (PDP 8), is
connected in a star pattern. It controls a total of
71 pumping stations by start/stop decisions, and
monitoring several variables.

Yoshizawa, : "Minicomputer Complex Systems", First USA- Japan Computer

K., Conference, 1972 , Session 15-4-1, pp. 480-485.

N., A highly technical description of a mini-network

Yoshida, Implemented at the Japan Racing Association for a

T., horse-betting information system. The network,

Yamada, which features "function-sharing" and "load-sharing",

E., and is shown to be superior to large computer systems in

Tanaka, T. both cost-performance and cost-reliability.


III. Distributed Data Bases

* Booth, : "The Use of Distributed Data Bases in Information

Grayce M. Networks", First International Computer Communication
Conference Proceedings . 1972, pp. 371-376.

Presents an overview of theorii?s concerning data base
creation and use within a computer network. Includes
alternative methods for creating a distributed data
base, how to match up jobs with the correct files,
problems of allowing a single application to access
files at several locations, and the problem of
protecting file and system integrity.

* Mantey, : "Integrated Data Bases for Municipal Decision-Making",

Patrick E. AFIPS . Vol. 44, 1975, pp. 487-493.
Carlson, Shows that, in a municipal setting, it is possible by

Eric D. means of "extraction" for "properly structured" files
on different computer systems to gain most of the
advantages of a data base maintained on a single
computer. These IBM authors therefore provide a case
which suggests that in some settings, one of the major
factors favoring large centralized computers, the
ability to maintain a large central data base, can be
handled just as well in the decentralized situation.

Rosenthal, : "The Distributed Data Base Concept", Guide 35 , 1972,
D.B. pp. 276-238.

Advocates distributed data bases in hierarchical systems.
Reviews pros and cons of both distributed and central-
ized data bases. Stresses the emergence of cost/
effective minis, and behavioral and reliability
advantages of distributing.

I V . Organizational Issues


* Argyris,

"Management Information Systems: The Challenge to
Rationality and Emotionality", Management Science ,
Vol. 17, No. 6, February 1971, pp. B-275 - B-292.

"If management informa
highest levels of aspi
conditions where execu
tion of space of free
failure and double bin
competence than formal
of essentiality. Thes
genuine resistance to
are not presently equi
problems caused by the
rationally "

tion systems achieve their designed
ration, they will tend to create
tives will experience (1) reduc-
movement, (2) psychological
d, (3) leadership based more on
power, and (4) decreased feelings
e experiences will tend to create
MIS. MIS specialists, in turn,
pped to cope with the emotional
ir systems. They react over

* Beckhard,

"Strategies for Large System Change", Sloan Management
Review, Winter 1975.

This article explores the topic of the successful
implementation of organization change in large systems.
Beckhard begins by describing a model of change planning
applicable to large and complex organizations. He then
focuses on five specific intervention strategies which
may be required in actual organization settings. The
article concludes with an examination of where in the
organization to begin a change effort and how to
maintain change once successfully initiated.

* Canning,
Richard G.

"Do We Have the Right Right Resources?", EDP Analyzer .
Vol. 13, No. 7, July 1975.

This report, the third in a four-part series on the
economics of data processing, includes a discussion
of the needs for thorough planning of the EDP activity,
and proposes a format for such a planning process.



Richard L,

"How to Control the Computer Resource", Harvard Business
Review, Nov. -Dec. 1973, pp. 68-78.

Includes an analysis of the differences between the
computer resource and other staff activities. Concen-
trates on the pricing mechanism of the EDP department,
comparing nonchargeout, full chargeout, any partial
chargeout. Suggests a framework for the design of the
charging method.


Ada B.

* Goggin,
William Z.

"Centralized versus Decentralized Computer Systems:

A New approach to Organizational Impacts", CISR working

paper. Number

This paper, based on the author's Ph.D. thesis, develops
a framework for analyzing the impacts of management
information systems on an organization. The framework,
based on the discipline of organizational development,
includes looking at planned change and at internal
organization dynamics. The paper includes, in addition,
an excellent summary of predictions identified in the
information systems literature concerning organizational
effects of centralization/decentralization.

"How the Multidimensional Structure works at Dow Corning",
Harvard Business Review , January-February 1974.

One emerging solution to the Centralization-Decentralization
issue is the matrix structure. This article describes
the implementation of a complex matrix structure at
jow Corning, it underlying philosophy, its use by manage
ment, and its advantages and disadvantages.


David L.


Ronald P.

"A Wholesale Retail Concept for Computer Network
Management", AFIPS , Vol. 41, Fall 1972, pp. 889-898.

Proposes a conceptual separation between the "wholesale
computing facility", whose task is to provide computing
power in several "grades", and the "retail computing
facility" whose task is to tailor applications to the
user's needs. This separation, say the authors, would
be beneficial in "identifying a viable management
structure for pooling computer resources across major
organizational boundaries" (into computer networks).

* Hebden,
J. E.

"The Importance of Organizational Variables in the
Computerization of Management Information Systems",
Journal of Management Studies , Vol. 8, No. 2, 1971,
pp. 179-198.

Discusses three organizational variables with direct
influence on the computer activity: technology, the
nature of the product and its market, and geographical
and social location of the organization. Concludes
with the predicted impact on EDP and the nature of
EDP staff and organization.


"Emerging EDP Patterns", Harvard Business Review ,
Vol. 48, March-April 1970, p. 16.

Details changes brought about by the computer in two


organizations. Changes in organizational structure
and process are described and analyzed. Includes
comparison with previous predictions and the author's
expectations for future changes.

Kolb, : "Organizational Development Through Planned Change:
David A. A Development Model", Sloan School of Management
and Working Paper 453-70, 1970.


Online LibraryJohn F. (John Fralick) RockartCentralization vs decentralization of information systems : an annotated bibliography → online text (page 1 of 2)