Modeling and Transforming Abstract Constraints for Automatic Service Composition

Automatic service composition typically consists of logical composition to produce an abstract workflow, followed by physical composition to transform it into a concrete workflow satisfying composition properties, such as Quality of service (QoS), preferences, and logic constraints. The composition properties have not been considered together up to now and, because they are abstract during the logical composition phase, they cannot be understood by the physical composer that reads concrete composition properties. This is an impediment to automatic service composition. Therefore, abstract properties in the logical composer have to be transformed to concrete properties for automated physical composition. This research investigates a stack of composition properties for QoS, preferences, and logic constraints considered together and architecture for automatic service composition, along with semi-automatic transformation of intermediate constraints in the architecture. The architecture together with an ontology for transforming constraints, and evaluation of our prototype are described.