3.2.

4 Algorithmic Cost Models systerm as the

software
estimated COst ofa system. Algo.
Algorithmic cost estimators compute the
that
comnprisethe
sum of the costs of the modules and subsystems model de
rithmic models are thus bottom-up estimators. algorithmic cost
(The Constructive Cost Model (COCOMO) is an summarized here. used t
COCOMO is briefly Section 3.1 are
seribed by Boehm BOE81 ). in
presented development schedulefor
(When using COCOMO, the equations
programnmer-months and
provide nominal estimates of Delivered Source Instructions
for product
of the estimate
program unit, based on the estimated number used to adjust Table
(DSI) in the unit. Effort multipliers are thenattributes, and project attributes
of values. The
attributes, computer attributes, personnel and their ranges from 63 soft-
3.4 summarizes the COCOMO effort multipliers by examining data
equations and effort multipliers were developed
the Delphi technique among a group of
ware development projects, and by use of
software experts. assumptions. For example,
a number of the
incorporate in
The COCOMO equations
the nominal organic mode (applications programs) equations apply following
lypes of situations:

Small to medium-size projects (2K to 32K DSI)

Familiar applications area
Stable, well-understood virtual machine
In-house development effort

Table 3.4 COCOMO effort multipliers

Range of values
Multiplier
Product attributes
0.75 to 1.40
Required reliability
Data-base size 0.94 to 1.16
Product complexity 0.70 to l.65
Computer attributes
Execution time constraint 1.00 to 1.66
Main storage constraint 1.00 to 1.56
Virtual machine volatility 0.87 to 1.30
Computer turnaround time 0.87 to 1.15
Personnel attributes
Analyst capability 1.46 to 0.71
Programmer capability 1.42 to 0.70
Applications cxperience 1.29 to 0.82
Virtual machine experience 1.21 to 0.90
Programming language experience 1.14 to 0.95
Project attributes
Use of modern programming practices 1.24 to 0.82
Use of software tools
1.24 to 0.83
Required development schedule 1.23 to 1.10
 Etfort mulipliers are used to modity these assumptions.
The following activities are covered by the estimates:

Covers design through acceptance testing

Includes cost of dovumentation and reviews
Inludes cost of projetmanager and program librarian
The effort estimators exclude planning and analysis costs, installation and
raining costs, and the cost of secretaries, janitors, and computer dperators. The
DSl estimate includes job control statements and source statements, but excludes
comments and unmodified utility routines. A DSI is considered to be one line or
card image. and a programmer-month consists of I52 programmer-hours.
Other assumptions concerning the nature of software projects estimated by
COCOMO include the following:
Careful definition and validation of requirements is performed by a small number
of capable people.
The requirements remain stable throughout the project.
Careful definition and validation of the architectural design is performed by asmall
number of capable people.
Detailed design, coding. and unit testing are performed in parallel by groups of
programmers working in teams.
Integration testing is based on early test planning.
intertace emors are mostly found by unit testing and by inspections and walk
throughs before integration testing
Documentation is performed incre mentally as part of the development process.
In other words, systematic techniques of software engineering are used
throughout the development process
COCOMO is
The following example of algorithmic cost estimation using
adapted from Boehm (BOE81 ):
product
The product to be developed is a 10-KDSI embedded mode software
microprocessor.
for telecommunications processing on a commercially available
The nominal effort equations for a 10-KDSlembedded mode product predict 44.4
programmer-months and 8.4 elapsed months for product development:
PM = 2.8*(10)**|.20 = 44.4
TDEV = 2.5*(44)**0.32 = 8.4
aspects of the
Effort multipliers are used to adjust the estimate for off-nominal
complex, but this is
project. For instance, the software is expected to be highly programmers. The
the planned use of highly qualified analysts and
Talble J.6 Cost cstimutlon proccdure using COCOMO.
I kdenuty all ubsyaten and molules in the product.
} Imate the size of cach module and caleulate the size of cach subsystem and the total
ystem
I Spey module level elfont multipliers for cach mnodule. The module-level multiplicrs
Ioduct complexity. pogrammner capability, virtual machinc cxperience, and programming
language expeielce.
4 Compue the module clfort and devclopment time cstimates for cach module, using the
nominal ctiDator cquations and the module-level cffort multipliers.
s Sncily the remaining || effort multipliers for cach subsystem.
onsteps and 5, compute the cstimatcd cffort and development time for each subsystem.
1 Irom step 6, conpute he total system cffort and development time.
8 erfom ascusitivity analysis on the cstimate to cstablish trade-off benefits.
Add other development costs, such as planning and analysis, that are not included in the
estimate.
rectify
l0. Compare the cstimate with one developed by top-down Delphi estimation. Identify and
the ditterences in the estimates

assumes that the

COCOMO is that use of a multiplicative effort adjustment factor
jIn reality, varying one factor often
various effort multipliers are independent. adjusted.
other factors should also be Often, it is not clear how
implies that
other factors.
variations in one factor influence the
version of COCOMO outlined here is Boehm's intermediate model. In his
The
presents a detailed model that accounts for differences in the effort
text, Boehm also
we have drawn heavily on Boehm's
multipliers by development phase. Althoughto present an exhaustive discussion of
text in thisdiscussion, our purpose is not influence software cost and the
indicate the factors that
COCOMO, but rather to
accounts for those factors.
manner in which COCOMO most cost estimation models do not
observation that
One final note is the code. In one instance, the cost of
reuse of existing
incorporate cost estimates for developing the equivalent amount of
new
cost of
reusing code was 20percent of the
code (ZEL83).