Composition-oriented Web Service Discovery

Web services are emerging as a promising technology for the development of next generation distributed heterogeneous software systems. Some prominent challenges for Web services are the discovery of services which match some given requirements, the composition of services into new complex applications, and the safe replacement of (sub)services. The primary objective of this thesis is to define a suitable technique for a semantics- and behaviour-aware service discovery, capable of satisfying complex client requests which specify the ontology-annotated inputs and outputs, and (possibly) the expected behaviour of the service to be found. We tackle the discovery, composition and replacement of services advertised by OWL-S descriptions, which provide a list of semantically annotated functional attributes of services, and a declaration of the interaction behaviour of services. First, we present a functional analysis that automatically generates (from a registry of OWL-S services) sets of services satisfying the functional requirements (viz., inputs and outputs) of client requests. Next, we describe a behavioural analysis that generates a (non-locking) composite service by suitably composing (the control-flow and the data-flow of) the services in a given set. The behavioural analysis checks whether the composite service satisfies the behavioural requirements of the query (viz., the expected service behaviour). To this end, a suitable notion of behavioural congruence for Web services is introduced. In order to show the applicability in practice of such a discovery technique, we also present a proof-of-concept implementation of the functional and the behavioural analyses, and we discuss possible approaches to the issues of scalability and service heterogeneity.