On Service-Oriented Architecture (SOA) construction projects, software architects are concerned with the characteristics of high quality services and how such services can be designed. For instance, they require advice regarding service interface design and guidance how to ensure that the constructed services meet the non-functional requirements that have been stated for them. To give such advice and guidance, service identification, specification and realization methods and techniques are required. Unlike service identification and specification, existing methods and techniques do not cover service realization well.

In our work, we investigate whether reusable architectural decision models can support service realization. Architectural decision models capture the knowledge (rationale) justifying certain designs. In the current state of the art, architectural decisions are captured ad hoc and retrospectively on each project, if at all; this is a labor-intensive undertaking without immediate benefits. On the contrary, we investigate the role reusable architectural decision models can play during SOA design: We treat recurring architectural decisions as first-class method elements and propose an architectural decision modeling framework and a reusable architectural decision model for SOA which guide the architect through the SOA design. Our approach is tool supported.

In the framework, we provide a technique to systematically identify recurring decisions. Our reusable architectural decision model for SOA conforms to a metamodel supporting reuse and collaboration. The model organization follows Model-Driven Architecture (MDA) principles and separates long lasting platform-independent decisions from rapidly changing platform-specific ones. The alternatives in a conceptual model level reference SOA patterns. This simplifies the initial population and ongoing maintenance of the decision model. Decision dependency management allows knowledge engineers and software architects to check model consistency and prune irrelevant decisions. Moreover, a managed issue list guides through the decision making process. To update design artifacts according to decisions made, we inject decision outcome information into design model transformations. Finally, a Web-based collaboration system provides tool support for the framework steps and concepts.

SOAD project news and results are available at: http://soadecisions.org/soad.htm