Substructure similarity search for engineering service-based systems

Abstract

With the broad application of service-oriented architecture in service-oriented software engineering, service-based systems (SBSs) are becoming ever more widely used. As a result, the selection of appropriate component services destined to fulfill the functional requirements becomes a critical challenge for successfully building SBSs, especially when the pre-specified SBS plan involves a complicated structure. Because building an exact SBS is often too restrictive, a similarity search for complex functional requirements becomes an essential operation that must be efficiently supported. We thus investigate in this work the substructure similarity search problem of building a SBS. To solve this new research problem, we propose a feature-based method, called the substructure similarity search for service-based systems (5S), to help users find similar SBS solutions by progressively relaxing a SBS plan. The 5S approach models each SBS as a set of features and transforms the task of relaxation of a SBS into the maximum allowed missing features, which can filter many SBSs directly without costly structure comparisons. 5S thus opens a new paradigm for efficient SBS engineering that shortens the build cycle. Finally, we discuss a series of experiments using real-world Web service datasets that demonstrate the effectiveness and efficiency of the proposed approach.