Publication

An essential driver for reducing Total Ownership Cost (TOC) and schedule is maintainability. This system quality (SQ) is key to reducing 75% of most systems’ life cycle costs. Also, Maintainability plays a key role in other top-level SQs: Life Cycle Efficiency, Dependability, and Changeability. Dependability needs Maintainability to relate Reliability to Availability; and Changeability needs Maintainability to address new system challenges and opportunities.

USC developed a tool called the Software Qualities Understanding by Analysis of Abundant Data (SQUAAD) for use in analysis of software technical debt1. SQUAAD has recently been extended to identify additional sources of technical debt by using uncompilability (computer code that fails to convert into machine instructions) as a symptom of careless development and analyzing software quality evolution over sequences of uncompilable commits. A commit is an event where computer code is modified, returned to the software code repository, and compiled (converted into machine instructions).

Another research focus was on the categorization for software commits and investigating how multiple-categories in a commit impacts software quality. Uncompilability increased when a software commit had more than two categories of change indicating a reduction in software quality. Work continues in refining categorization and training models to automate categorizing commits.

Research was also done on the identification of synergies and conflicts between system qualities. Using computer code from open-source Apache projects and the CAST tool, a correlation analysis was done to detect synergistic and conflicting qualities across ten different quality metrics. This research can provide a better understanding of what trade-offs will need to be made and the costs associated with ensuring levels of different qualities for the intended system.

A new security cost driver has been developed for COCOMO II and for later use in the emerging definition of COCOMO III. This involved surveys of the Linked-In security community, presentation and discussion of the proposed security cost driver rating scale at the SERC Doctoral Forum, and an experts’ workshop at the 2019 annual COCOMO Forum, and an experts’ Delphi consensus of the cost multipliers for each of the rating scale levels. This work will result in understanding how the total cost of a software system can be reduced with the appropriate allocation of resources in the early stages of the software project, thus reducing TOC.