CEN 4021 Software Engineering II Effort estimation Instructor: Masoud Sadjadi http://www.cs.fiu.edu/~sadjadi/ sadjadi@cs.fiu.edu CEN 4021 20th Lecture Acknowledgements Dr. Onyeka Ezenwoye Dr. Peter Clarke CEN 4021: Software Engineering II 20th Lecture 2 Agenda Effort estimation – General Estimation Model – COCOMO Model CEN 4021: Software Engineering II 20th Lecture Effort Estimation Software cost and effort estimation will never be an exact science. There are too many parameters, e.g., human, technical, environmental, and political parameters, that affect the actual estimation making it almost impossible to have considered all cases. Even if all parameters are considered, the following questions still need to be answered: CEN 4021: Software Engineering II 20th Lecture Effort Estimation 1. 2. How much does each parameter contribute to the total effort estimation? How can the amount of each parameter’s contribution be converted into some numerical form? 3. How do the parameters affect each another? 4. How would these effects be combined? The above point to the fact that there is a lack of good metric definitions related to estimations. CEN 4021: Software Engineering II 20th Lecture Effort Estimation Historically, managers have based many of their estimates on some form of “consulting the expert”: – Consult other peer managers with experience – Engage consultants with experience – Draw analogies to past projects that seem to have similar characteristics – Divide and conquer – break project into components and get estimates for each component based on experience, then combine the estimates The practice of consulting the experts and using past experience has worked well in the past CEN 4021: Software Engineering II 20th Lecture General Estimation Model It is important to have as accurate an estimation as possible, since this estimate is used in planning the schedule and resources. Several researchers have constructed effort and cost estimation models. To date these models have produced mixed results. Software effort estimation model: A set of information and relationships organized for the purpose of estimating the effort needed to complete a s/w project. The information and the values assigned to the information may vary from project to project. CEN 4021: Software Engineering II 20th Lecture General Estimation Model Software effort estimation model cont: Similarly, the relationships may be organized into a mathematical equation, which may vary from project to project. Most models use the following general form, or its derivative, to estimate effort: Effort = (a + [b x (Sizec)]) x PROD(f’s) where a, b, c = statistically derived coefficients or best approxs. Size = estimated size of the project f’s = factors that influence the project estimates PROD(f’s) = the product of arithmetically multiplying those factors CEN 4021: Software Engineering II 20th Lecture General Estimation Model The coefficients a, b, c were derived by fitting the best curve against known data for the Effort and Size parameters for previous projects. Effort is usually measured in terms of person-months expended on the project. Size is usually the number of lines of code created as part of the project. The Size metric is itself a problem! If it is measured in terms of “lines of code” (LOC), then there must be clear definition and agreement on LOC. CEN 4021: Software Engineering II 20th Lecture General Estimation Model How do you define LOC? – Should LOC include only executable code? – What about documentation of code? – Is an LOC written in assembler the same an LOC in C++ or JavaScript? – How do we account for the difference between an LOC in assembler language code and an LOC in an HLL? CEN 4021: Software Engineering II 20th Lecture General Estimation Model Other Factors affecting Project Effort The effort required for a project also depends on the conditions under which the project is being undertaken. These conditions include among others: – Programmer productivity – S/w support services e.g., levels of testing – Maturity of the organization, working with an established process. – Type and amount of tool support CEN 4021: Software Engineering II 20th Lecture General Estimation Model All the preceding factors, the f’s in the general estimation equation, affect the project. It is a good idea to have a standard list of factors to be considered during estimation. This list will: – serve as a reminder – bring some consistency to what the manager consider The list of factors (f’s) may also be interrelated. The question of how to account for the combined effects of all these factors continues to be a perplexing challenge. CEN 4021: Software Engineering II 20th Lecture The COCOMO Effort Estimation Model The COnstructive COst Model (COCOMO) is a specific example of the general estimation equation developed by Boehm in the 1980’s. COCOMO includes 3 levels of models: 1. A macro estimation model 2. An intermediate-level model 3. A more detailed micro estimation model The intermediate model is used in the example in the text. CEN 4021: Software Engineering II 20th Lecture The COCOMO Model General process of using the COCOMO is: 1. Choose an estimate of what would be considered the “nominal” (typical) development of the project. Three nominal project modes are defined: Organic, Semidetached, and Embedded. 2. Choose an estimate of the size of the project to use the Size parameter. 3. Review the factors that influence the project, the cost drivers, and estimate the influence that each factor will have on the chosen “nominal” case. CEN 4021: Software Engineering II 20th Lecture The COCOMO Model General process of using the COCOMO is cont: 4. Determine the effort for the s/w project by inserting the estimated values into the Effort Formula. Nominal Mode of Development The following key project characteristics are used to identify one of the three nominal models: A. The teams understanding of the project objective B. The teams experience with similar or related projects C. The project’s need to conform with the estimate requests CEN 4021: Software Engineering II 20th Lecture The COCOMO Model D. The project’s need to conform with the established external interfaces E. The need to develop the project concurrently with new systems and new operational procedures F. The project’s need for new and innovative technology, architecture, or other constraints G. The project’s need to meet or beat the schedule H. The project’s size range Organic mode – projects are fairly easy and familiar to the s/w development team. CEN 4021: Software Engineering II 20th Lecture The COCOMO Model Semidetached mode – projects are medium sized and some what familiar to the s/w development team. Embedded mode – complex projects that may be unfamiliar to the project team. The above summary statements should be used as a starting point in the decision process when choosing the most appropriate mode for the project estimate. See Table on next page (Table 17.1 in text) CEN 4021: Software Engineering II 20th Lecture Table 17.1 Mode of the Development Use in COCOMO Key project charac. Organic mode Semidetached mode Embedded mode A Detail degree Considerable degree Only general degree B Extensive amount Some amount None to modest amount C Only the basic ones Considerably more than the basic ones All and full conformance D Only the basic ones Considerably more than the basic ones All and full conformance E Little to some Moderate amount Extensive amount F None to minimal Some Considerable G Low Medium Must H < 50,000 delivered LOC 50,000 to 30,000 delivered LOC All sizes CEN 4021: Software Engineering II 20th Lecture The COCOMO Model An untrained manager might not be able to easily choose the correct mode. Very few projects have characteristics that fall neatly within any one of the mode categories. For each of the three modes, a different formula is used for estimating the Effort: Organic: Effort (in person-months) = 3.2 x (Size)1.05 Semidetached: Effort (in person-months) = 3.2 x (Size)1.12 Embedded: Effort (in person-months) = 3.2 x (Size)1.20 Size is expressed as LOC. CEN 4021: Software Engineering II 20th Lecture Assigning value to cost drivers Category Product attributes Factor Range values Very low low Nominal High Very high Extra high RELY 0.75 0.98 1.0 1.15 1.40 - DATA - 0.94 1.0 1.08 1.16 - 0.85 1.0 1.15 1.30 1.65 CPLX Computer attributes TIME 0.70 - 1.0 1.11 1.21 1.66 STOR - - 1.0 1.06 1.30 1.56 VIRT - 0.87 1.0 1.15 1.15 - TURN - 0.87 1.0 1.07 0.71 - ACAP - 1.19 1.0 0.86 0.82 - AEXP 1.46 1.13 1.0 0.91 0.70 - PCAP 1.29 1.17 1.0 0.86 - - VEXP 1.42 1.10 1.0 0.90 - - LEXP 1.21 1.07 1.0 0.95 0.82 - MODP 1.14 1.10 1.0 0.91 0.83 - TOOL 1.24 1.10 1.0 0.91 1.10 - SCED CEN 4021: Software Engineering1.23 II 1.08 1.0 1.04 Personnel attributes Project attributes 20th Lecture Calculate the effort estimate Organic: Effort = [3.2 x (Size)1.05]xPROD(f’s) Semidetached: Effort = [3.2 x (Size)1.12]xPROD(f’s) Embedded: Effort = [3.2 x (Size)1.20]xPROD(f’s) CEN 4021: Software Engineering II 20th Lecture