Methods for Efficient and Accurate Discovery of Services

The discovery of services is of one of the most integral parts of a service-oriented system. Discovery is the problem of identifying services from a pool of service descriptions that fulfill the requirements of a discovery request. With an increasing number of services developed and offered in an enterprise setting or the Web, users can hardly verify their requirements manually in order to find the appropriate services. Automated discovery methods can support users in discovering appropriate services that they have not been aware of before. The ability to discover services effectively depends on how services are advertised, how requirements can be expressed, and how the requirements are verified. It is challenging to develop service discovery methods that can be applied in a wide variety of use cases, while providing a good trade-off between expressivity and efficiency. In this thesis, we develop a method to discover semantically described services. The discovery method exploits comprehensive service and request descriptions that capture functional and non-functional properties. In our discovery method, we compute the matchmaking decision by employing an efficient model checking technique. Our logic-based discovery method automatically identifies accurate matches for a given request. The proposed method can be applied to services that describe their complete behavior in the form of executable process expressions. In addition, we introduce an alteration of the method tailored to discover services that cannot disclose their complete behavior and provide an interface description instead. In order to facilitate service discovery in large bodies of offered services, we propose approaches for more efficient matchmaking. Formal service classes enable an automated and consistent service classification and induce a class hierarchy, which can be utilized as an offline index structure. While each class in the index is described by a given request, we also propose indexing structures that can be automatically populated either offline or online, i.e., during the processing of incoming requests. The offline indexes accelerate reasoning tasks by materializing possible propositions in advance. The online index aims at caching frequent requests such that repetitive requests can be processed faster. Our contributions are based on scenarios from current fields of research and have been implemented and evaluated in the context of largescale research projects.