Integrated Schedule Risk Analysis in Software Development

1. Schedule Risk Analysis of Software Development By: Dr. Joe H. Dean Engineering Chief Risk & System Analysis LFWC And: Mr. David W. Benson, Jr. Engineering Specialist Avionic Software Engineering LFWC BV41840

2. Current Practice – Schedule Risk Analyses for Software Projects Are Typically Performed for Individual Software Tapes However, No Quantification of Integrated Schedule Risk Is Made Short Coming – If a Tape Is High Risk (or Moderate or Low) What Does That Mean Regarding the Probability of Timely Completion Of That Tape? Of Effected Programs? Needed – Cumulative Analysis to Show Sensitivity of Integrated Program Schedules to Tasks on Individual Software Tapes Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data Why Integrated Schedule Risk Analysis? BV41841

3. Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data Schedule Risk Model Basis & Capability Risk Issues Program Schedules • Task Duration • Consequences • Abatement Options Min. Most Max Likely Task Relationships SS FS FS FS SS FF FS FF SF Schedule Risk Analysis Task Relationships: Robustness SS Start-to-start Sensitivities SF Start-to-finish Delivery Dates FS Finish-to-start Prob SS Finish-to-Finish Time • Program Schedule Risk • Alternative Critical Paths • Abatement Analysis BV41842

4. Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data FS: Finish-to-Start Task Connectivity Within & Between Tapes Is Modeled Software Start TAPE 1 FS FS FS FS FS FS ... System Design Detailed Design Preliminary Design Code/Unit Test CSC Test BV41843

5. Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data FS: Finish-to-Start Task Connectivity Within & Between Tapes Is Modeled (Cont’d) Software SW Start TAPE 1 FS FS FS FS FS FS ... System Design Detailed Design Preliminary Design Code/Unit Test CSC Test FS FS FS TAPE 2 FS FS FS FS ... System Design Detailed Design Preliminary Design Code/Unit Test BV41844

6. Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data Task Connectivity Within & Between Tapes Is Modeled (Cont’d) Task Relationships FF: Finish-to-Finish SF: Start-to-Finish FF: Finish-to-Finish SF: Start-to-Finish Software Start FS FS FS FS FS FS System Design Detailed Design Code/Unit Test CSC Test Tape 1 (Prog. 1) Preliminary Design FS FS FS FS FS FS FS System Design Preliminary Design Detailed Design Code/Unit Test Tape 2 (Prog. 1) FS FS FS FS FS Tape 3 (Prog. 1) Detailed Design Preliminary Design System Design FF FF FF Tape 4 (Prog. 1) Preliminary Design Detailed Design FF Tape 1 (Prog. 2) System Design Preliminary Design Detailed Design FS FS FS Tape 2 (Prog. 2) Preliminary Design FS Tape 1 (Prog. 3) System Design Preliminary Design FS LOA FS FS Tape 2 (Prog. 3) Preliminary Design Flight Test BV41845

7. Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data REVIC Basics • COCOMO (Constructive Cost Model) Developed by Dr. Barry Boehm Based on Curve Fit of Approx. 70 Projects Across Industry 75% OF NOMINAL • Model Calibration via 15 Factors Covering Product, Computer, Personnel, and Project Attributes Software Task (e.g. DSI) • Model Covers Tasks from Availability of Firm Software Requirements (Assumed Reasonably Stable) through Completion of Software Test NOMINAL RISK • Schedule Compression Adds Cost and Risk and Cannot Be Compressed to Less than 75% of Nominal Time • REVIC (Revised Version Intermediate COCOMO) Based on USAF Specific Projects • The F-16 SPO and LFWC Have Used COCOMO/REVIC Since the Mid 1980’s BV41846

8. Task Completion Variations Are Assigned Based on REVIC Analysis Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data High Risk Moderate Risk Nominal Risk Schedule Less than 80% of Nominal Schedule 80 - 92.5% of Nominal Schedule Within 92.5% of Nominal 90% 100% 140% 90% 100% 120% 90% 100% 110% 75% OF NOMINAL Task Size (DSI) NOMINAL RISK Time BV41847

9. Monte Carlo Process In Schedule- Risk Model Is Straightforward Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data 1 2 Enter Task Completion Variability and Connectivity Between Tasks Random Draw Determines Completion Time for Each Task 1 Random Number FS MIN MAX Completion Time FF SS FS 0 Min Most Likely Max 3 4 For Each Trial (Total Network), Construct Critical Path & Project Completion Date. At Any Given Time (Date), Probability of Success # of Trials Complete Total # Trials = [Repeat N Trials] 1 Trial Results P (t) S 0 0 Time (Completion Dates) t = Time BV41848

10. Probability of Completion vs. Planned Finish Time Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data 1.00 0.75 Probability of Completion 0.50 0.25 0.00 -60 -40 -20 0 20 40 60 80 100 Number of Days from Planned Finish PLANNED TIME (BASELINE) BV41849

11. Assessing Schedule Changes Resulting from Risk Abatement Plan Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data 1.00 RISK ABATEMENT ALTERNATIVE 0.75 Change In Probability Completion BASELINEPLAN Probability of Completion 0.50 0.25 Change In Planned Finish 0.00 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 Number of Days from Planned Finish BV41850

12. Results Corresponded Well with Expert Opinion Software Task/Tape Completion Times Were “Reasonably” Bounded in Regard to Program Schedule Impact Customer Needs Analysis & Control Requirements Allocations Verification Design Product Data Summary Preliminary Findings Appraisal • This Process Is a Promising Method to Quantify Schedule Uncertainty, Especially in Highly Interrelated Task Networks • This Process Allow Potential Schedule “Savings” from Risk Abatement Alternatives to Be Considered as an Aid in Decision Making BV41851