Adaptation Goals for Adaptive Service-Oriented Architectures

Service-oriented architecture supports the definition and execution of complex business processes in a flexible and loosely-coupled way. A service-based application assembles the functionality provided by disparate, remote services in a seamless way. Since the architectural style prescribes that all features be provided remotely, these applications adapt to changes and new business needs by selecting new partner services to interact with. Despite the success of the architectural style, a clear link between the actual applications —also referred to as service compositions— and the requirements they are supposed to meet is still missing. The embedded dynamism also imposes that requirements properly state how an application can evolve and adapt at runtime. The solution proposed in this chapter aims to solve these problems by extending classical goal models to provide an innovative means to represent both conventional (functional and non-functional) requirements and adaptation policies. To increase the support to dynamism, the proposal distinguishes between crisp goals, of which satisfiability is boolean, and fuzzy goals, which can be satisfied at different degrees; adaptation goals are used to render adaptation policies. The information provided in the goal model is then used to automatically devise the application’s architecture (i.e., the composition) and its adaptation capabilities. The goal model becomes a live, runtime entity whose evolution helps govern the actual adaptation of the application. All key elements are exemplified through a service-based news provider. Luciano Baresi and Liliana Pasquale Politecnico di Milano, Dipartimento di Elettronica e Informazione, piazza L. da Vinci, 32 20133 Milano Italy, e-mail: {baresi|pasquale}@elet.polimi.it ∗ This research has been funded by the European Commission, Programmes: IDEAS-ERC, Project 227977 SMScom, and FP7/20072013, Projects 215483 S-Cube (Network of Excellence).