Software Size Estimation Copyright, 2006 © L. Ouyang Liubo Ouyang [email protected] Personal Software Process Lecture 5.

Software Size EstimationSoftware Size Estimation

Copyright, 2006 © L. Ouyang

Liubo OuyangLiubo Ouyang

[email protected]@126.com

http://ss.hnu.cn/oylb/psp/http://ss.hnu.cn/oylb/psp/

Personal Software Process Personal Software Process

Lecture 5Lecture 5

L.Ouyang, PSP, Lecture 5

IntroductionIntroduction

What to estimate software size for?What to estimate software size for?

• To make plans. To make plans.

• To estimate the effort. To estimate the effort.

• To create a reliable schedule.To create a reliable schedule.

““Sizing represents the project Sizing represents the project planner’s first major challengeplanner’s first major challenge””

-- R.S. Pressman-- R.S. Pressman



COCOMO:COCOMO:

• Effort = a Effort = a SizeSizebb

COCOMO 2.0:COCOMO 2.0:

• Effort = (Effort = (eeii) ) A A SizeSizeBB

Walston-Felix model:Walston-Felix model:

• Effort = 5.2 Effort = 5.2 SizeSize0.910.91

Software equation (Putnam):Software equation (Putnam):

• Effort = (Effort = (SizeSize B B0.3330.333/P)/P)33/ t/ t44



.. Project planning .... Project planning ..

Size estimatingSize estimating


Plan of the lecturePlan of the lecture

• Delphi methodDelphi method

• Fuzzy-logic methodFuzzy-logic method

• Standard-component methodStandard-component method

• Function point methodFunction point method

• Probe methodProbe method



• Delphi methodDelphi method• Fuzzy-logic methodFuzzy-logic method• Standard-component methodStandard-component method• Function point methodFunction point method• The Probe methodThe Probe method


Wideband-Delphi methodWideband-Delphi method

Rand Corporation, Boehm’81Rand Corporation, Boehm’81

• A few experts individually A few experts individually produce size estimates.produce size estimates.

• A Delphi process is used to A Delphi process is used to reach a consensus. reach a consensus.

PythiaPythia



1. Experts get the 1. Experts get the specificationspecification and an and an estimation formestimation form

2. They meet for 2. They meet for discussiondiscussion (project goals, (project goals, assumptions, estimation issues)assumptions, estimation issues)

3. Each expert 3. Each expert anonymouslyanonymously lists the lists the taskstasks and and estimatesestimates the size the size

4. The estimates go to the 4. The estimates go to the estimate estimate moderatormoderator. He tabulates the results and . He tabulates the results and returns them to the experts.returns them to the experts.

The Delphi procedureThe Delphi procedure

The estimateThe estimate

moderatormoderator



Estimator: Jerzy Nawrocki Date: 22.06.1999Estimator: Jerzy Nawrocki Date: 22.06.1999

Project: Sorting routineProject: Sorting routine

The estimates from the 1st round:The estimates from the 1st round:

e e EE MM e e e e

0 20 40 60 80 1000 20 40 60 80 100

e - estimates, e - estimates, EE - - your estimateyour estimate, , MM - - median estimatemedian estimate

Your estimate for the next round: ......... LOCYour estimate for the next round: ......... LOC

A rationale for your estimate: ............................................A rationale for your estimate: ............................................

........................................................................................................................................................................................



5. The experts meet to discuss the results. 5. The experts meet to discuss the results. They review the They review the taskstasks they have defined they have defined but but notnot their their size estimatessize estimates..

6. The procedure is repeated from step 3 6. The procedure is repeated from step 3 until the estimates are acceptably nearuntil the estimates are acceptably near

The Wideband-Delphi procedureThe Wideband-Delphi procedure

The estimateThe estimate

moderatormoderator



• Delphi methodDelphi method

• Fuzzy-logic methodFuzzy-logic method• Standard-component methodStandard-component method• Function point methodFunction point method• Probe methodProbe method


Fuzzy-logic methodFuzzy-logic method

Putnam’92:Putnam’92:

• We need We need accurateaccurate, not , not necessary necessary preciseprecise size size estimates.estimates.

• Let’s relate the estimate Let’s relate the estimate to the to the historical datahistorical data

Tall Middle ShortTall Middle Short



Range Low Middle High

V. small 11 000000 22 000000 44 000000

Small 44 000000 88 000000 1166 000000

Medium 1166 000000 3322 000000 6644 000000

Large 6644 000000 112288 000000 225566 000000

V. large 225566 000000 551122 000000 11 002244 000000

Fuzzy-logic size rangesFuzzy-logic size ranges



Given the smallest (Given the smallest (SS) and the ) and the largest (largest (LL) program size, find ) program size, find range boundaries A, B, C, D range boundaries A, B, C, D such that such that SS, A, B, C, D, , A, B, C, D, LL form form a a geometrical progressiongeometrical progression..

A/A/SS = B/A = C/B = D/C = = B/A = C/B = D/C = LL/D = p/D = p

LL//SS = p = p55 p = ( p = (LL//SS))0.20.2

If If SS= 1 000 and = 1 000 and LL= 1 024 000, = 1 024 000, then p=4then p=4

V. Large: D - V. Large: D - LLV. Large: D - V. Large: D - LL

Large: C - DLarge: C - DLarge: C - DLarge: C - D

Medium: B - CMedium: B - CMedium: B - CMedium: B - C

Small: A - BSmall: A - BSmall: A - BSmall: A - B

V. small: V. small: SS - A - AV. small: V. small: SS - A - A



Remarks:Remarks:• Historical data are Historical data are

required on a large required on a large number of projects.number of projects.

• Don’t change the existing Don’t change the existing ranges. Add new ranges ranges. Add new ranges when required.when required.

Tall Middle Short ? Tall Middle Short ?



• Delphi methodDelphi method• Fuzzy-logic methodFuzzy-logic method

• Standard-component Standard-component methodmethod

• Function point methodFunction point method• Probe methodProbe method


Standard-component methodStandard-component method

Putnam’92Putnam’92

• We have a system We have a system decomposed into a set of decomposed into a set of standard components standard components (screens, files, modules (screens, files, modules etc.)etc.)

• For each standard For each standard component we have component we have historical data about its sizehistorical data about its size


Standard-component methodStandard-component method

You have to provide for each component type:You have to provide for each component type:

SS - the smallest number of components of that - the smallest number of components of that type;type;

LL - the largest number - the largest number

MM - the most likely number - the most likely number

Estimated number of components of a given type:Estimated number of components of a given type:

((SS+4*+4*MM++LL) / 6) / 6

For each component type we know average LOC.For each component type we know average LOC.



• Delphi methodDelphi method• Fuzzy-logic methodFuzzy-logic method• Standard-component methodStandard-component method

• Function point methodFunction point method• Probe methodProbe method


Function point methodFunction point method

Albrecht, IBM, 1979Albrecht, IBM, 1979

Basic functions:Basic functions:

• InputsInputs

• OutputsOutputs

• InquiriesInquiries

• Internal data filesInternal data files

• External interfacesExternal interfaces



DesignedDesigned

applicationapplication

InternalInternaldata filedata file

OutputOutput

InputInput

InquiryInquiry

AnotherAnother

applicationapplication

ExternalExternalinterfaceinterfacefilefile



An An outputoutput: a report, a screen, an : a report, a screen, an error message. Individual data error message. Individual data items within a report are not items within a report are not counted separately.counted separately.

An An inquiryinquiry: an on-line input that : an on-line input that results in an on-line output. An results in an on-line output. An inquiry cannot modify any internal inquiry cannot modify any internal file.file.

Output Output Inquiry Inquiry



Fun types Simple Aver. Complex Total

Inputs x 3 x 4 x 6

Outputs x 4 x 5 x 7

Inquiries x 3 x 4 x 6

Intern. files x 7 x 10 x 15

Interfaces x 5 x 7 x 10

Unadjusted total

2 2 2 26

3 3 3 48

2 1 0 10

2 1 0 24

0 1 0 7

115

ProblemProblem: simple, average or complex?: simple, average or complex?



FP = UT * CMFP = UT * CM

FP - Function pointsFP - Function points

UT - Unadjusted total (previous slide)UT - Unadjusted total (previous slide)

CM - Complexity multiplier (0.65 .. 1.35)CM - Complexity multiplier (0.65 .. 1.35)

CM = 0.65 + 0.01 * CM = 0.65 + 0.01 * Influence_factors Influence_factors

14 influence factors, 0 - 5 points each14 influence factors, 0 - 5 points each



Influence factor evaluationInfluence factor evaluation

0 - No influence0 - No influence

1 - Incidental1 - Incidental

2 - Moderate2 - Moderate

3 - Average3 - Average

4 - Significant4 - Significant

5 - Essential5 - Essential



Are Are data communicationsdata communications required? required?

Are there Are there distributed processingdistributed processing functions?functions?

Is Is performanceperformance critical? critical?

Will the system run in a Will the system run in a heavily used heavily used operational environmentoperational environment??

Does the system require Does the system require on-line data on-line data entryentry??

Is the internal Is the internal processing complexprocessing complex??

Is the code designed to be Is the code designed to be reusablereusable??



Are the inputs, outputs, files or Are the inputs, outputs, files or inquiries inquiries complexcomplex??

Does the on-line data entry require Does the on-line data entry require the input the input transactiontransaction to be built to be built over over multiple screensmultiple screens or or operations?operations?

Are the master Are the master files updated on-linefiles updated on-line??

Are conversion and Are conversion and installationinstallation included in the design?included in the design?



Does the system require reliable Does the system require reliable backup and recoverybackup and recovery??

Is the system designed for Is the system designed for multiple multiple installationsinstallations in different in different organisations?organisations?

Is the application designed to Is the application designed to facilitate facilitate change and ease of usechange and ease of use by the user?by the user?



LanguageLanguage

Assembly languageAssembly language

CC

CobolCobol

FortranFortran

PascalPascal

AdaAda

o-o languageso-o languages

spreadsheetsspreadsheets

LOC/FPLOC/FP

320320

128128

105105

105105

9090

7070

3030

66



Modifications of the basic FP model:Modifications of the basic FP model:

• FeatureFeature points (C. Jones, 1991) points (C. Jones, 1991)

• ObjectObject points (R. Kauffman, R. Kumar, points (R. Kauffman, R. Kumar, 1993)1993)

• 3D function3D function point (S.A. Whitmire, 1995) point (S.A. Whitmire, 1995)



• Delphi methodDelphi method• Fuzzy-logic methodFuzzy-logic method• Standard-component methodStandard-component method• Function point methodFunction point method

• Probe methodProbe method


Probe methodProbe method

Humphrey, CMU, 1995Humphrey, CMU, 1995

PROPROxy-xy-BBased ased EEstimatingstimating

Objects as proxies Objects as proxies


Planning a software projectPlanning a software project

Conceptual designConceptual design

Size estimates Size estimates

Resource estimates Resource estimates

The schedule The schedule

The product The product

RequirementsRequirements

Size databaseSize database

Productivity databaseProductivity database

Resources availableResources available

Size, res., sched. dataSize, res., sched. data



1. Prepare a conceptual 1. Prepare a conceptual design (objects and design (objects and methods + their methods + their function)function)



2. For each method 2. For each method assign one of size assign one of size rangesranges

Very big Big Medium Small Very smallVery big Big Medium Small Very small



3. Knowing:3. Knowing:• programming languageprogramming language• object categoryobject category• size rangessize ranges• the number of methodsthe number of methods

estimate, using historical estimate, using historical data, size of each data, size of each object.object.



4. Determine initial 4. Determine initial program estimated program estimated size, X, adding the size, X, adding the values received in the values received in the previous step.previous step.

2 + 3 = 52 + 3 = 5



5. Apply linear 5. Apply linear regression to get regression to get estimated program estimated program size Y:size Y:

Y = Y = 11 X + X + 00

5 means 105 means 10



xxi i yyii - n x - n xavgavg y yavgavg

xxii22 - n x - n xavgavg

22 11 = =

00 = y = yavgavg - - 11 x xavgavg



6. Using the 6. Using the t distributiont distribution and and standard standard deviationdeviation compute the compute the prediction intervalprediction interval for for a given percentage. a given percentage.

For 100% theFor 100% the

interval isinterval is

[0; + [0; + ]]



6a. Calculate the standard deviation, 6a. Calculate the standard deviation, , of , of the regression function.the regression function.

6b. To find the two-sided value of 6b. To find the two-sided value of tt for the for the probability probability qq, look in a table of the , look in a table of the tt distribution under distribution under p(p()= (1+q)/2)= (1+q)/2 and and n-2n-2 degree of freedom.degree of freedom.


The t distributionThe t distribution

Degrees p()=.6 p()=.7 p()=.85 p()=.9

5 .267 .559 1.156 1.476

6 .265 .553 1.134 1.440

7 .263 .549 1.119 1.415

8 .262 .546 1.108 1.397

9 .261 .543 1.100 1.383



(x(xkk - x - xavgavg))22

(x(xii - x - xavgavg))22++

11

nn++11 Range = t Range = t

6c. Compute the range as follows:6c. Compute the range as follows:


SummarySummary

After requirements analysis:After requirements analysis:• Function pointsFunction points

After high level design:After high level design:• Wideband-Delphi methodWideband-Delphi method• Fuzzy-logic methodFuzzy-logic method• Standard-component Standard-component

methodmethod

After low level design:After low level design:• Probe methodProbe method


Further readingsFurther readings

Function Point Counting Practices Manual Function Point Counting Practices Manual Release 4.0, IFPUG, http://www.ifpug.org.Release 4.0, IFPUG, http://www.ifpug.org.

How Are Function Points Useful, Longstreet How Are Function Points Useful, Longstreet Consulting, http://www.SoftwareMetrics. Consulting, http://www.SoftwareMetrics. com/manyuser.htm.com/manyuser.htm.

R.Pressman, Software Engineering - A R.Pressman, Software Engineering - A Practitioner Approach, McGraw-Hill, New Practitioner Approach, McGraw-Hill, New York, 1997 (4th edition), Ch. 4.3York, 1997 (4th edition), Ch. 4.3

W. Humphrey, A Discipline for Software W. Humphrey, A Discipline for Software Engineering, Addison-Wesley, Reading, Engineering, Addison-Wesley, Reading, 19951995


Quality assessmentQuality assessment

1. What is your general 1. What is your general impression ? (1 - 6)impression ? (1 - 6)

2. Was it too slow or too fast ?2. Was it too slow or too fast ?

3. Did you learn something 3. Did you learn something important to you ?important to you ?

4. What to improve and how ?4. What to improve and how ?

Software Size Estimation Copyright, 2006 © L. Ouyang Liubo Ouyang [email protected] Personal Software Process Lecture 5.

Documents

e e e e

e estimates

size estimating slide

size beffort

ouyang liubo ouyang

fuzzylogic method putnam92

estimate moderator slide

size b cocomo