A Dependency Based Impact Analysis Framework for Service-Oriented System Evolution

Service-oriented architecture (SOA) has become a widely accepted approach that provides a flexible IT infrastructure in order to deal with the increasing pace of business changes and global competition. However, the evolution of SOA challenges traditional research methodologies and motivates us to explore new approaches for analyzing and evaluating the effects of change. In order to identify the potential change effect to a software system, the impact analysis can be important to assist the system maintainers to make evolution decisions.;To support our proposed framework, we have developed an impact analysis tool that allows change analysts to identify service dependencies, make changes, and to perform impact analysis. We have used this tool to evaluate the actual industry Web Services for service synchronization.;Keywords. Service-Oriented Architecture, Web Services, Software Evolution, Impact Analysis, Software Dependency, Service Synchronization;In this thesis, we first present a formal framework for analyzing the dependencies, the changes, and the impact of these changes to service-oriented system evolution. Specifically, we identify service dependencies at inter- and intra- service levels and develop graph-based service dependency matrices. Based on the calculation of the matrices, the ripple effect that indicates dependencies and the impact of the change on the service can be analyzed and quantitatively measured. Additionally, by extending the generic impact analysis framework, we have analyzed service synchronization through a version dependency model and function-tree dependencies. We have also performed three types of ripple effect analysis for facilitating the transition of client applications. Furthermore, based on the dissipative structure, dependency link analysis, and information entropy, we have proposed an approach for examining the service-oriented system evolution. The dependency links are used to represent service dependencies and the underlying information models shared by these service dependencies. Our approach analyzes the architecture impact caused by the service dependency changes and the model correspondence changes. According to the case study, our proposed approach provides valuable insights for examining the effects of the system changes.