1. Copyright 2000, Stephan Kelley1 Estimating User Interface Effort Using A Formal Method By Stephan Kelley 16 November 2000

2. Copyright 2000, Stephan Kelley2 Outline Problem Introduction to Software Estimation Introduction to COCOMO Approach Schedule Further Research

3. Copyright 2000, Stephan Kelley3 Problem Software Estimation has not included the modern man-machine interface as a complete, but separate, part of radio communcation systems. Consequently, the user interface is not properly budgeted into radio system designs, resulting in hasty, incomplete implementations near the project deadline.

4. Copyright 2000, Stephan Kelley4 Problem Modern user interfaces are complex Graphical User Interfaces (GUIs) are composed of more software than their predecessors Today’s GUI allows the system user access to more functionality than was previously available

5. Copyright 2000, Stephan Kelley5 Objective Break out the system’s user interface estimation into a separate function within system software estimation. In this way, the user interface software’s cost and effort are more accurately predicted before implementation.

6. Copyright 2000, Stephan Kelley6 Objective Separate Graphical User Interface (GUI) effort estimation should reduce the extent to which the GUI contributes to a project being behind schedule and over budget.

7. Copyright 2000, Stephan Kelley7 Previous work Sparse - GUI software estimation is not a thoroughly researched topic. General software estimation methods have been developed by Watts S. Humphrey (CMU), Capers Jones/A.J. Albrect, Barry Boehm (USC), Larry Putnam, and others

8. Copyright 2000, Stephan Kelley8 Definition Estimation - Function of predicting the size of a software project and the amount of effort and the cost required to complete the project. Estimates are generated by various models from real - usually historical - data.

9. Copyright 2000, Stephan Kelley9 Estimation Methods Function Points (Jones/Albrect) - estimation by measuring system functionality SLIM (Software LIfe Cycle Management - Larry Putnam) - estimation via statistics, simulation, and extensive reviews of the software project PROBE (PROxy-Based Estimating - Humphrey)

10. Copyright 2000, Stephan Kelley10 Estimation Methods cont’d COCOMO (Constructive COst MOdel - Boehm) - Estimation by projected system size and inclusion of project-specific cost factors.

11. Copyright 2000, Stephan Kelley11 Drawbacks Function Points - oriented toward data processing applications and results can be interpreted subjectively. SLIM - not suitable for small projects; sensitive to fast software project buildup times. COCOMO - sensitive to historical data and fast software project buildup times.

12. Copyright 2000, Stephan Kelley12 Method COCOMO is the method of choice for the following reasons Easy to use Various factors affecting software development are transparent to the engineer The objective is to demonstrate a concept, not evaluate the method

13. Copyright 2000, Stephan Kelley13 COCOMO COCOMO is a simple software cost and effort estimation model based on the relationships among the number of source instructions and four coefficients which depend on the software project’s type and another parameter called the Effort Adjustment Factor (EAF).

14. Copyright 2000, Stephan Kelley14 COCOMO cont’d COCOMO II is a revision of COCOMO released in 1997 by the Center for Software Engineering at Southern California University with successive releases in 1998, 1999, 2000, and one planned for 2001.

15. Copyright 2000, Stephan Kelley15 COCOMO cont’d Each successive version of COCOMO has been calibrated to a greater number of sample projects (called data points) and used a more rigorous approach to combine empirical data with opinions of developers to calibrate the latest version model’s parameters.

16. Copyright 2000, Stephan Kelley16 COCOMO cont’d The original COCOMO method generates the following estimation data: Effort (man months): PM = EAF * a * size b Duration (months): T dev = c * PM * d Staffing: P = PM/T dev

17. Copyright 2000, Stephan Kelley17 COCOMO coefficients Each coefficient has three different values from which one is chosen for a project based on whether the project is organic, semi- attached, or embedded. Development Mode a b c d Organic3.21.052.50.38 Semi-Attached3.01.122.50.35 Embedded2.81.202.50.32

18. Copyright 2000, Stephan Kelley18 COCOMO coefficients Embedded - the application operates on small margins of error and interfaces extensively with hardware; large Organic - the project involves small development teams familiar with the application domain; often small Semi-attached - a fuzzy category between embedded and organic projects

19. Copyright 2000, Stephan Kelley19 COCOMO coefficients Since a radio is a potentially large, hardware- intensive, real-time application, the embedded mode will be used.

20. Copyright 2000, Stephan Kelley20 COCOMO coefficients Effort Adjustment Factor (EAF) - product of 15 software cost drivers which will be multiplied with the previously computed devlopment time value T dev

21. Copyright 2000, Stephan Kelley21 EAF Cost Drivers Required reliability Database size Product complexity Execution time Requirements Main storage constraints Virtual machine volitility Computer turnaround time Analyst capability Applications experience Programmer capability Virtual machine experience Programming language experience Programming practices Software tools use Development schedule

22. Copyright 2000, Stephan Kelley22 Assumption A sufficient requirements analysis has occurred to allow valid software development effort estimation.

23. Copyright 2000, Stephan Kelley23 Approach Use the COCOMO software effort estimating method to determine the cost and effort needed to adequately implement the graphical user interface to a complex radio system.

24. Copyright 2000, Stephan Kelley24 Approach cont’d Select four projects from supporting industry. Two will attempt to estimate the cost and effort required for the user interfaces using the COCOMO method; the user interface parts of the other two projects will not be estimated for cost and effort. For each project, record the number of hours spent implementing the user interface.

25. Copyright 2000, Stephan Kelley25 Approach cont’d At the end of the experiment time frame, the difference between the actual effort spent on the user interface and the projected effort will be calculated. For the “unestimated” projects, the projected effort will be obtained from system engineering, which normally calculates effort required on the project.

26. Copyright 2000, Stephan Kelley26 Approach cont’d The difference values for the estimated projects will be compared with the difference values for the “unestimated” projects. Analysis of project post-mortems will also be made to determine what factors may have contributed to the difference between the GUIs’ effort estimations.

27. Copyright 2000, Stephan Kelley27 Schedule Can be difficult to evaluate method because the typical project lifecycle is longer than one year. Must be flexible since start of experiment is determined by start of host projects’ start dates and deadlines.

28. Copyright 2000, Stephan Kelley28 Schedule cont’d Obtain approval of four projects for observation - approx. 1 month Duration of project experiment - 10 months Compile and analyze data; report - 1 month

29. Copyright 2000, Stephan Kelley29 Facilities No special computing facilities will be necessary; supporting industry will provide Aid of industry partners is highly beneficial

30. Copyright 2000, Stephan Kelley30 Further Research Compare effort and cost estimations for GUI implementation using another, more rigorous method Compare effort and cost estimations for GUI implementation for other types of systems besides radio communications