A Demand-driven, Dynamic Evolution Of The Model And Method-oriented Software Architecture View,

With the deepening and development of software engineering, the importance and universality of software evolution are improving. The changing process of software systems to meet the requirements of users is software evolution, which can be divided into static evolution and dynamic evolution, and this thesis focus on dynamic evolution. At present, the dynamic evolution is a focus and nodus of research in academia, for one thing, the dynamic software evolution which has become an important form of software life cycle is a key field of software engineering research, and for another, the research of software dynamic evolution still faces many challenges and difficulties. Therefore, aim at challenges which dynamic software evolution faces, the research on how to improve dynamic evolvability of software systems to enhance the reliability of the implementation of software dynamic evolution attach importance to the theoretical and immediate significance.Aim at challenges which dynamic software evolution faces, this thesis holds this opinion:solutions of those challenges can not be sought from one isolated angle, and they should be set about during the whole software life cycle, meanwhile it is indispensable that software dynamic evolution should be supported by several important stage of software process. Therefore, this thesis focus on establishing a systematic modeling theory of dynamic software evolution with software requirement model as a driven, software architecture model as a view, behavior monitor mechanism as a support, addressing the crucial challenge of dynamic evolution as a objective, and formal method as the basic means. For this purpose, there main progresses have been made in this thesis:(1) A dynamic-evolution-oriented software requirements meta-model was designed. To be started, feature was selected as basic components, and an approach of modeling software requirements based on the combination of features was offered. Requirements model was composed of behavior feature and property feature, furthermore, behavior feature was divided into calculation behavior feature and interaction behavior feature to isolate interaction from calculation. After that, through mapping the scope of property feature model to the range of behavior feature model, behavior feature model and property feature model were integrated as the ultimate requirements model.(2) The method of normalizing dynamic-evolution-oriented software Requirements model was proposed. Through the divide strategy, behavior feature models and property feature models were normalized respectively. For behavior feature models, the normal form of the process algebra term was designed, and it was proofed that each term (if meet rules) can be converted into normal form by axiom systems. On this basis, for property feature models, referential integrity, dependency consistency and mutex consistency of property feature models were discussed, and property feature models were normalized as first normal form to fourth normal form.(3) A dynamic-evolution-oriented software architecture meta-model was designed. The meta-model consists of two views:static views and dynamic views. For static views, the static structure of software systems were described. The basic components of static views includes:components and connectors. Extended Petri nets structure were selected to modeling components, while connectors include three basic connecting ways: palaces integration, transitions integration and arcs adding. For dynamic views, they were established based on static views, tokens of extended Petri nets were added into static views, after that, the dynamic behavior of software systems can be described.(4) The method of transforming requirements models into architecture models was proposed. To be started, behavior equivalence was selected as the basis of models transforming. After that, rules of transforming basic components in requirements model into counterparts in architecture models were proposed. Further, rules of transforming combination components were also proposed. Finally, the structural properties of architecture models which transformed from requirements models were analyzed.(5) Dynamic-evolution-oriented behavior monitor mechanism was proposed. To be started, the concept of behavior monitor was given, on this basis, three functions of behavior monitor were discussed, including management, monitoring, and evolution supporting. For management, three aspects of management including token management, palaces management and transitions management were discussed. For monitoring, three aspects of monitoring including the state of components management, the target accessibility of token, the mutex control of places and transitions were discussed. For evolution supporting, several important algorithms was designed, including:driven component to static state, driven component to active state, add component, delete component, add connector, delete connector. (6) An approach of components relevance analysis was proposed based on behavior monitor mechanism. To be started, the structural relevance between components were analyzed, and structural relevance can be devided into basic structural relevance and composition structural relevance.Further, on the basis of structural relevance analysis, the part of transitivity in behavior relevance was discussed. Finally, the algorithms of components behavior relevance analysis in closed-systems and in open-systems were designed based on behavior monitor mechanism.(7) The standards of consistency preserve during the process of component dynamic evolution implementation were proposed. To be started, the state migration methods of components were discussed. After that, the formal definition of behavior space and behavior graph was given as the basis of consistency preserve theory. Further, References strong simulation and weak simulation from process algebra theory, the standards of consistency preserve, including internal consistency preserve and external consistency preserve, were proposed. Finally, it was mathematical proof that the standards were correct and reasonable.In summary, this thesis provides a deep research in software dynamic evolution. And the work of this thesis will improve the efficiency and quality of the implementation of software dynamic evolution by reducing cost and time in the process.