An integrated approach for improving the performance of semantic service discovery

Service Oriented Computing (SOC) promotes the development of business applications through service composition where more than one atomic services which are capable of realizing the given business goal are discovered semantically and composed in a specific pattern. Though semantics has brought automation and dynamism into composition, semantic service discovery process is time consuming. Clustering techniques are extensively used to improve the performance of semantic discovery, as partitioning services by a desired characteristic lays the foundation for many of the service mining tasks. Though clustering technique reduces the search space of discovery from pool of all available services to a specific cluster, it does not eliminate the invoking of semantic reasoning completely as each service of the specific cluster is still in need of semantic reasoning to detect whether the service is matched one or not according to inputs and outputs of the query. Hence, in this paper an integrated approach which combines clustering and indexing techniques is proposed to avoid the overhead of invoking semantic reasoning during querying. The approach clusters the services by their output parameters with desired threshold similarity. An input index is maintained for each cluster(that resulted in clustering) which readily produces a list of matched services without invoking reasoning during querying. Time characteristics of the proposed approach are analyzed using typical test services and the results are presented. The integrated approach is found to yield better performance improvement.