Quantifying software architecture attributes
Department of Computer Science
Doctor of Philosophy
Leung, Joseph Y-T.
Shih, Frank Y.
Klashner, Robert Michael
Product line architecture
Software architecture holds the promise of advancing the state of the art in software engineering. The architecture is emerging as the focal point of many modem reuse/evolutionary paradigms, such as Product Line Engineering, Component Based Software Engineering, and COTS-based software development.
The author focuses his research work on characterizing some properties of a software architecture. He tries to use software metrics to represent the error propagation probabilities, change propagation probabilities, and requirements change propagation probabilities of a software architecture. Error propagation probability reflects the probability that an error that arises in one component of the architecture will propagate to other components of the architecture at run-time. Change propagation probability reflects, for a given pair of components A and B, the probability that if A is changed in a corrective/perfective maintenance operation, B has to be changed to maintain the overall function the system. Requirements change propagation probability reflects the likelihood that a requirement change that arises in one component of the architecture propagates to other components. For each case, the author presents the analytical formulas which mainly based on statistical theory and empirical studies. Then the author studies the correlations between analytical results and empirical results.
The author also uses several metrics to quantify the properties of a Product Line Architecture, such as scoping, variability, commonality, and applicability. He presents his proposed means to measure the properties and the results of the case studies.
njit-etd2006-043 (147 pages ~ 8,106 KB pdf)
Please complete this Feedback Form to inform us about your experience using this website. It will assist us in better serving your information needs in the future. Thank You!
Created September 8, 2008