Building dependable service-oriented application via dynamic reconfiguration and fault-tolerant reconfiguration collaboration protocol

Software development techniques are now moving rapidly towards Service-Oriented Computing paradigm. A Service-Oriented Application is to build application with services that can be looked up, published and bound at the execution time across boundary of organizations. One major concern of the service-oriented mission-critical system is to deliver dependable services. Traditionally, building application with dynamic reconfigurability is one of the major approaches to promote system dependability. However, building dependable service-oriented application is challenging because it faces an open, heterogeneous computing environment, where numerous distributed services perform computations concurrently and collaboratively. The negotiation between application and service for reconfiguration planning and selection is difficult due to the rapidly increasing number of available services in an ad-hoc environment. Existing dynamic reconfiguration systems are not adequate to face this challenge.; This research presents a dynamic application/service recomposition and reconfiguration methodology with fault-tolerant distributed reconfiguration protocol to build dependable service-oriented application. The application reconfiguration selection is based on a reconfiguration ontology and the accompany reconfiguration selection and reasoning algorithm. The domain specific reconfiguration ontology enables the negotiation of application and service for reconfiguration in a more effective and collaborative manner. The reconfiguration goes beyond the existing reconfiguration approach of service rebinding by developing a distributed reconfiguration consensus protocol. It takes into account the service and application dependency relations that exist in many of today's applications. The reconfiguration consensus protocol determines whether or not a reconfiguration can be accepted by all the participants for a reconfiguration event. This behavior guarantees that no uncoordinated reconfiguration actions could break application integrity. Further, it has been recognized that without a dependable underlying reconfiguration framework, the application and services it reconfigures are not truly dependable. Thus, the consensus protocol and distributed agents that carry out the reconfigurations are designed to be fault-tolerant themselves. A scratch proof is provided for the correctness of the algorithm.