“COCOMO II : MODEL FOR ESTIMATING SOFTWARE COST”

PRESENTED TO
DR. BAHRAM KHALILI
AS PART OF
HONORS CREDIT CONTRACT

CSE 3310, FALL 2008,

KAPIL VYAS,
CSE, 1000502888

i
 TABLE OF CONTENTS

1. ABSTRACT......................................................................................................3

2. HISTORY .........................................................................................................4
2.1 Boehm, Barry: ........................................................................................................................ 4
2.2 Cocomo Background: ............................................................................................................ 4

3 CURRENT STATUS ......................................................................................6

3.1 Available Computerized Editions of Cocomo II:..................................................................... 6

4. ARCHITECTURAL OVERVIEW ......................................................................7

4.1 Applications Compositional Model ......................................................................................... 7
4.2 Early Design and Post-Architecture Models .......................................................................... 8

5. PAPER COMPARISON .................................................................................10

5.1 Early Design Model: ............................................................................................................. 10
5.2 Post Architecture Model:...................................................................................................... 10
5.3 A Cross-Comparison Between The Two Models:................................................................ 10

6. ADVANTAGES AND DISADVANTAGES ......................................................12

6.1 Significant Aspects of Cocomo II: ........................................................................................ 12
6.2 What is Being Worked On? ................................................................................................. 12

7. CONCLUSION AND FUTUREWORK............................................................13

REFERENCES....................................................................................................14
1. Abstract

This article introduces the reader to the constructive cost model (COCOMO)
II – a well known model used in software cost and schedule estimation. First,
the background of the model is presented, tracing the history of COCOMO
from its inception to the present time. We then briefly examine its current
status. Next, we present a moderately-detailed architectural overview.
Thereafter, advantages and disadvantages of making use of the model are
described shortly. We then compare two research papers in the area and
finally conclude with future work suggestions.
2. History

2.1 BOEHM, BARRY:

One of the most prominent contributions of Barry is the Constructive Cost
Model (COCOMO) used in estimating Software cost. COCOMO II is a well
known model used in software cost and schedule estimation. Barry
Boehm is TRW professor of software engineering and Computer Science
Department Director. He received his BA degree from Harvard in 1957
and his MS and PhD degrees from UCLA in 1961 and 1964, all in
mathematics. His current research interest include software process
modeling, software requirements engineering, software architectures,
software metrics and cost models, software engineering environments,
and knowledge-based software engineering. Besides his COCOMO
contributions to the field, others are the Spiral Model of the software
process, the Theory W (win-win) approach to software management and
requirements determination, and two advanced software engineering
environments: the TRW Software Productivity System and Quantum Leap
Environment.

2.2 COCOMO BACKGROUND:

Significant research on software cost modeling began with the extensive
1965 SDC study of the 104 attributes of 169 software projects (Nelson,
1966). This led to some useful partial models in the late 1960s and early
1970s. In the late 1970s, several sophisticated software cost estimating
models developed in response to the demand for control of escalating
software costs. The proprietary RCA PRICE-S model, now the PRICE-S
model, was released in 1977 and was instantly successful (Park, 1988). A
year later, Lawrence Putnam’s Quantitative Software management Company
marketed the first edition of the proprietary Software Life Cycle Model (SLIM),
which has also been successful (Putnam and Myers, 1992). Barry W. Boehm,
then employed by TRW Corporation, was considering the same approach: to
develop a proprietary algorithmic (or parametric) software cost model to be
released to the public, much like PRICE-S and SLIM. However, Boehm
(1981) decided to use a different approach; he wrote the book Software
Engineering Economics, which completely described and explained
COCOMO.

COCOMO quickly became popular because it was not proprietary, free to

use, and relatively easy to learn and operate. COCOMO user’s groups
emerged throughout the world, and several computerized editions soon
became available. As the years passed COCOMO’s popularity increased but
the model experienced difficulties in estimating software projects of the 1990s
as a result of new practices such as non-sequential and rapid-development
process models; reuse-driven approached involving commercial-off-the-shelf
(COTS) packages, reengineering, applications composition , and application
generation capabilities; object-oriented approaches supported by distributed
middleware; software process maturity effects and process-driven quality
estimation. There was a need to develop COCOMO II to meet the changing
needs of the 1990s and the twenty-first century. COCOMO II was an effort
that was started to update software cost estimation models, such as the 1981
Constructive Cost Model and its 1987 Ada COCOMO (Boehm and Royce,
1987; Boehm, 1989) update.

Every year the COCOMO User’s Group meets at the University of Southern
California chaired by Boehm who is the director of the Center of Software
Engineering. At these meetings Boehm and the rest of the COCOMO II
research team (in alphabetic order – Chris Abts, Jongmoon Baik, A. Winsor
Brown, Sunita Chulani, Brad Clark and others) usually unveil COCOMO
enhancements, some of which are discussed later in this paper. Also, there
are several computerized editions, some of which are discussed later in this
paper.
3 Current Status

The USC CSSE overlooks the progress of COCOMO II ever since it came
into existence. Their website at http://csse.usc.edu/csse boasts around 42
industry and government affiliates who help them to keep current with the
technology and economic trends through their annual workshops and
collaborative projects. Their affiliates include Cisco, Boeing, IBM,
Samsung, Microsoft, General Dynamics, DARPA, Motorola and Sun
Microsystems among others. They also have courses wherein 5-student
teams go from inception through to transition in 24 weeks, developing
real-client e-service applications for campus (and some off-campus)
organizations. They use a mix of industrial-grade tools (Rational
Rose/Soda/Clear Case, MS Project, numerous Java and other
development tools) and our own USC-developed tools (USC COCOMO II,
Easy Win-Win) and the aforementioned methods. The website mentions
of their upcoming 2009 CSSE-Annual Research Review at the USC
Campus from March 16 to 19. The next event is International Forum on
COCOMO and Systems/Software Cost Modeling on October 26 to 29 in
Washington D.C. Based on these facts there is evidence that these
models are heavily used in industry and there is on-going work to
continuously improve the model while maintaining to the changing needs
of the time.

3.1 AVAILABLE COMPUTERIZED EDITIONS OF COCOMO II:

The three most widely used COCOMO II tools are Costar, Cost Xpert, and
Estimate Professional.

Costar is a commercial tool sold by Softstar Systems. It contains all COCOMO

updates, from the COCOMO 81 to the Ada COCOMO and now COCOMO II, and
provides many useful tables and graphs. This commercially available tool is the
most widely COCOMO tool used. More information can be obtained at
http://www.softstarsystems.com

Cost Xpert is a commercial tool sold by Marotz, Inc. It supports seven different
costing methodologies in addition to COCOMO II and Feature Points. More
information of Cost Xpert can be found at http://www.marotz.com

Estimate Professional is a tool sold by Software Productivity Centre, CA. It is a

commercial management tool that combines the most well-known and reliable
estimation models, COCOMO II and Putnam’s SLIM, with Monte Carlo
simulation. For more information, please visit http://www.spc.ca.
4. Architectural Overview

The COCOMO II models were developed to support the future software

practices marketplace as described in (Boehm et al., 1995). The three
main models [described in detail in the COCOMO II Model Definition
Manual (University of Southern California, 1997, 2000) and summaries of
which are presented below] are:
1. Applications Composition
2. Early Design
3. Post-Architecture

4.1 APPLICATIONS COMPOSITIONAL MODEL

A rapidly increasing number of projects are being developed using
Applications Composition (model) methods that rely on having an
integrated computer-aided software engineering (ICASE) environment. In
such environments, source lines of code (SLOC) and function points
(FPs) are not a good measure of the software size due to their limitations
of sizing systems developed using GUI builders and visual programming
aids.

The COCOMO II team found a highly attractive metric called Object

Points, developed by (Banker et al., 1994) for estimation of ICASE-based
financial applications. Object Points (OP) uses the number of screens,
reports, and third-generation language (3GL) modules as basic sizing
primitives. The function point convention of weighting these by complexity
and adding them together to obtain an overall size metric, which they then
adjusted for reuse, is used. The COCOMO II team adopted OPs counting
rules with two changes: (1) added rating scales for determining a project’s
productivity rate in NOP/PM, where NOP stands for New Ops, in terms of
the ICASE system’s maturity and capability, and the developers’
experience and capability in using the ICase environment; and (2)
changed the name from Object Points to Application Points, to avoid
confusion with a number of sizing metrics for more conventional object-
oriented applications.

Figure 1 presents the baseline COCOMO II Application Point procedure

for estimating the effort involved in Application Composition and
prototyping projects.
• NOP: New Object Points (Object Point count adjusted for reuse)
• Srvr: number of server (mainframe or equivalent) data tables used
in conjunction with the SCREEN or REPORT
• Clnt: number of client (personal workstation) data tables used in
conjunction with the SCREEN or REPORT
• %reuse: the percentage of screens, reports, and 3GL modules
reused from previous applications, prorated by degree of reuse.

The Applications Composition model has been calibrated to the 19 data

points from Banker et al. giving an accuracy of estimates within 30% of the
 actuals less than 50% for the full sample. However, the percentage of
estimates within a factor of 2 of the actuals is 89% for the full sample. This
appears reasonable to associate a factor of 2 with the optimistic and
pessimistic range limits for an Application Point estimate. The COCOMO II
team has not been able to collect much further data on applications
composition projects for calibration and hence classifies this model in its
emerging extensions of COCOMO II.

4.2 EARLY DESIGN AND POST-ARCHITECTURE MODELS

This subsection presents the Post-Architecture and Early Design models. The
Post-Architecture model is a detailed model, and as the name suggests, it is
used once the architecture of the system is well defined, namely, when the
product is ready to develop and sustain a fielded system. Detailed information
on cost driver inputs must be available so accurate cost estimates can be
made. The Early Design model is a high level model that is used in the
exploration of architectural alternatives or incremental development
strategies. This level of detail is consistent with the general of estimation
accuracy needed. The Post-Architecture and Early Design models use the
same sizing approach and the same scale drivers. To simplify the flow of the
discussion, these common modeling features will be presented first followed
by a description of the Post-Architecture effort multipliers and then the Early
Design effort multipliers.

The effort equation for the Post-Architecture and Early Design models is as
follows:

PM = A * Size^(E) * Sum of product of EMi from i = 1 to n

Where E = B + 0.01* sum of SFj from j =1 to 5

PM = effort in person-months
A = baseline multiplicative constant
B = baseline exponential constant
EM = effort multiplier
SF = scale factor

For the Post-Architecture model, n = 17 and for the Early Design model, n =
7. The SFs are the same for both the models. The 2000 calibration of the
models gave A = 2.94 and B = 0.91.

The schedule equation for the Post-Architecture and Early Design models is:
TDEV = C * (PMns)^F
Where F = D + 0.2*(E-B)
Here:
TDEV = schedule in time for development in calendar months
C = baseline multiplicative constant
D = baseline exponential constant
PMns = effort (i.e., PM) without the effects of required development
schedule ( the SCED effort multiplier, PMns stands for PM with nominal
schedule. The 2000 calibration of the models gave C = 3.67 and D = 0.28)
Sizing is one of the main inputs for the Post-Architecture and Early Design
models of COCOMO II. Determining the size of product includes determining
new code written, code reused from other sources, with or without
modifications, and automatically translated code (SLOC) and/or unadjusted
function points (IFPUG, 1994), which are then converted to SLOC. The
Software Engineering Institute (SEI) definition checklist for a logical source
statement is used in defining the LOC measure ( Park, 1992; Goethert et al…
1992) The effect of reuse needs to be accounted for in the sizing metric and
is done by the nonlinear model.
5. Paper Comparison

Here we compare 2 papers that discuss the Early Design and Post-
Architecture Models.

5.1 EARLY DESIGN MODEL:

This model is intended mainly for architectural design activities. It uses
Unadjusted Function Point (UFP) counts which are converted to DSI.
Application adjustment factors are not applied until after conversion to SLOC.
effort, PMNS = a x Size b x Π EMi (i = 1 to 6)
where a = 2.94 (calibrated from 161 projects)
b = 1.01 + 0.01 x Σ SFj (j = 1 to 5)
and EMi - effort multipliers for 6 different cost drivers
SFi - exponential scale factors for 5 cost drivers
NS - implies nominal schedule

5.2 POST ARCHITECTURE MODEL:

This model is intended mainly for actual development activities. It is essentially a
modern update to COCOMO 81. It considers a wide range of drivers. New cost
drivers include reusability, documentation needs, personnel continuity and multi-
site teams.

effort, PMNS = a x Size b x Π EMi (i = 1 to 16)

where a = 2.94 (calibrated from 161 projects)

b = 1.01 + 0.01 x Σ SFj (j = 1 to 5)

and EMi - effort multipliers for 16 different cost drivers

SFi - exponential scale factors for 5 cost drivers
NS - implies nominal schedule

5.3 A CROSS-COMPARISON BETWEEN THE TWO MODELS:

• Size:
 o Early design and Post-Architecture both make use of Function
Pont (FP) with language or SLOC
• Reuse:
o Early design and Post-Architecture have Equivalent SLOC
equal to nonlinear f(AA, SU, UNFM, DM, CM, IM)
• Requirements Change:
o Both make use of REVL (requirements evolution)
• Maintenance:
o Both use f(MCF, SU, UNFM) where MCF = (%added +
%modified)
• Scale drivers, B
o Both have the same equation: B = 0.91 – 1.23 depending on
degree of precedentedness, development conformity, early
architecture and risk resolution, team cohesion, and process
maturity.
• Product cost drivers
o Early design makes use of RCPC(a,b) and RUSE(a,b)
o Post architecture makes use of RELY(a,b), DATA(a),
DOCU(a,b), CPLX(a,b), RUSE(a,b).
• Platform Cost Drivers
o In early design platform difficulty is given by PDIF(a,b) while in
Post model we use TIME(a), STOR(a), PVOL(a)
• Project Cost Drivers:
o Early design makes use of CED(a) and FCIL(a,b)
o Post-architecture makes use of TOOL(a,b), SCED(a) and
SITE(a,b).
6. Advantages and Disadvantages

6.1 SIGNIFICANT ASPECTS OF COCOMO II:

Some of the many significant aspects of COCOMO II include: replacement of the
3 modes in COCOMO 81 with a single exponential equation for effort estimation
and one for schedule estimation, five scale factors for adjusting the exponents of
the equations, three sizing options, redefined and additional cost drivers, a
requirements evolution and volatility factor, a non-linear reuse model, two levels
of cost-driver granularity, phases and milestones for three types of development
processes, and a Bayesian calibration method. Some of the major aspects are
listed below:
Aspect Elaboration
Five exponential scale factors precedentedness, development
flexibility, architecture/risk resolution,
team cohesion, process maturity
Adaptation Adjustment Multiplier non-linear cost of assessment and
assimilation, software understanding,
and unfamiliarity for reuse of software

Three sizing options application points, function points,

Three levels of cost model granularity
Three development processes
New cost drivers
source lines of code
Application Composition, Early Design,
Post-Architecture
waterfall, MBASE, incremental
software reuse, required
documentation, personnel continuity,
and multiple development sites

Bayesian calibration historical data plus expert judgment

6.2 WHAT IS BEING WORKED ON?

Recent work on COCOMO methodology at the USC Center for Software
Engineering has focused on developing extensions to and new models based on
COCOMO II, as well as work to unify the various models into a single,
comprehensive estimation tool. A COCOMO II extension is an estimation model
that uses the output of COCOMO II and modifies it in various ways. A new model
is one based on the COCOMO approach to model building but which requires its
separate inputs. New models can be used in conjunction with COCOMO II if
desired. Information about COCOMO II extensions and new models can be
found on the web site for the USC Center for Software Engineering.
7. Conclusion and Future work

COCOMO II, its extensions and new COCOMO-like models incorporate many
innovations intended to accommodate a variety of modern approaches to
software and software-intensive systems. This paper aimed at looking at the brief
history of how the model came about, the current computerized editions that
implement the model by various vendors. We then saw the architectural view
discussing the three main classification models: Applications Composition model,
early design model, and Post-architecture model. We eventually move onto
looking at a detailed comparison between the seeming similar early design
model and post-architecture model. In the end we look at significant aspects of
COCOMO II that has made it the most popular model for software cost
estimation. The next goal is to unify the various models into a single
comprehensive estimation tool.
References

Boehm, Barry, et al., Software Cost Estimation with COCOMO II , Prentice-

Hall, 2000. ISBN 0-13-026692-2.
Marciniak, John J., Encyclopedia of Software Engineering, Volume 1, 2nd
Edition, John Wiley & Sons Inc., New York 2002, Pg 79, 150-163
Stutzke, Richard, Estimating Software-Intensive Systems Projects, Products
and Process, Pearson Education Inc., Upper Saddle River, NJ. April 2005
Fairley, Richard E.. "The Influence of COCOMO on Software Engineering
Education and Training". IEEE Xplore. November 12th, 2008
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=01617347

B. Boehm, Software Engineering Economics, Prentice-Hall, Englewoods Cliff,

NJ, 1981.
Boehm, C. Abts, A. W. Brown, S. Chulani, B. Clark, E. Horowitz, R. Madachy,
D. Reifer, and B. Steece, Software Estimation with COCOMO II, Prentice
Hall, Upper Saddle River, NJ, 2000.
University of Southern California, COCOMO II Model Definition Manual, USC,
Center for Software Engineering, Computer Science Department, Los
Angeles, CA, 1997. Available at website:
http://sunset.usc.edu/COCOMOII/cocomo.html

Schett, Nancy M., Seminar on Software Cost Estimation, University of Zurich,

Switzerland, 2002-2003. Pg. 13