As the development of the software technology, the software system has graduallyevolve step by step from the traditional static, reactive software system to open,dynamic, self-adaptive and service-oriented software systems. The demand of systemruntime requirement management and self-adaptive decision-making method become animportant issue.Base on the common understanding to the basic concepts of the running systems, Iproposes a conceptual framework to illustrate the shift of paradigms from conventionaldesign-time requirements analysis, to run-time reaction according to the environmentalchanges, to today’s on-demand services requirements and capability match-making andcomposition. The framework point out what has to be defined prior to running, and whatis changeable during run time, how to make the system adaptable to run-time demandswithout taking the system offline. A goal oriented adaptive method is proposed tosupport the conceptual framework.In this thesis, I analysis the service based adaptive system and give out the conceptabout the atom service, service set, service flow and the service sequence. Base on therun-timestate and the quality of service, I evaluate the satisfaction of the running servicesequence of the system. The system apply the Markov decision making process to thesystem self-adaptive process. The system define the four basic elements of decisionmaking process. System state is defined by the running service sequence, system actionsare defined by the self-adaptive action, the transform probability is based on thesequence availability.The system would evolve when the system is running based on theself-adaptive decision making process. During the process, the adaption rules are alsogenerated by the system. Furthermore a bookstore order handling system is used toillustrate the availability of the method.In addition, a browser based system analysis and decision making modeling tool isdesigned and implemented to support the proposed method. |