Dynamic Evolution-Oriented Component Lifecycle Control Research

The dynamic evolution of component systems is currently the research hotspot of the component technology. Existing approaches to realize the dynamic evolution usually apply ad hoc procedures and mechanisms, as the lifecycle of the components is defined coarsely and the internal statuses that are essential to the control of dynamic evolution process are intangible. This paper proposes extended component lifecycle models, which define the "safe" status toward different dynamic evolution protocols. A lifecycle control technique is also introduced based on the reflective extension on component model. Users/administrators of the system could guide a component to reach a safe point for the evolution through a set of explicit management interfaces. The main contributions of this paper are summarized as following:1) Targeting various consistency assurance protocols for component dynamic evolution, including Quiescence, Tranquility, and Version-consistency, we propose extended component lifecycle models. Each model refines the lifecycle of the component with an additional safe status and the precise semantics of state transitions are stated.2) The design principles of the supporting framework for component lifecycle control and the lifecycle controller are introduced. Based on component reflection techniques, we extend the component model and design the static dependency graph, transaction machine and dynamic transaction dependency list at component meta-level. We also make a detailed description on the working principles and how to support system consistency control.3) A new programming model of the components with lifecycle callback methods that are triggered during the dynamic evolution process is proposed. With this model, lifecycle callback methods are declared in the definition of a component by developers. And when a component reaches some specific lifecycle status during the evolution, customizable actions can be taken at component level to cooperate with the system.4) A reference lifecycle control system for dynamic evolution of EJBs is implemented. The user can observe the statues of components and guide a component to reach a safe point (Quiescence, Tranquility, Version-consistency) for dynamic evolution through a set of management interfaces.