Deriving the Cost of Software Maintenance for Software Intensive Systems

Abstract

Throughout software's lifetime, changes are introduced to the code in order to maintain the desired performance. These changes often create side effects, which cause other cascading effects elsewhere in the software or other system components with which the software interfaces. In a sense, the software degrades because of the maintenance performed on it, not because of a lack of maintenance upkeep. This pattern makes the cost of software maintenance difficult to predict, given the amount of variability in the upkeep process. Therefore, the best that program managers can hope for are heuristics that permit them to approximate annual operating budgets when calculating total ownership costs. Typically, these methods employ metrics used during development to estimate the annual cost of maintaining the software (i.e., source lines of code or function points). Through correlation and regression analysis, this thesis examines 62 programs that captured software maintenance data to determine a cost model for software maintenance. Even though a model was not built, the main contribution of this thesis is to provide a greater awareness of the complexity of estimating the costs for software maintenance. Additionally, this thesis provides insight to cost variables that may assist program managers when estimating annual software maintenance costs.