An ontology-based modeling methodology for service-oriented architecture

Abstract

With the rapid growth of the software industry, Service-Oriented Architecture (SOA) has been considered as a new paradigm for system development and integration. By using services to encapsulate functionalities of business tasks and providing standard communication between services, SOA provides a design framework for realizing rapid and low-cost system development and improving total system quality. SOA modeling is the initial phase of SOA development lifecycle and the quality of the SOA model will directly affect the quality of SOA application. Applying ontology techniques to SOA modeling can provide accurate descriptions for models, identify the binding information of business process and service, increase the reusability of existing business processes and services, and accelerate application development. In this thesis, we adopt ontology techniques to assist SOA modeling, developing a core ontology BPO (Business Process Ontology) for business process modeling as well as proposing an ontology-based SOA application modeling and developing framework. BPO can provide accurate definitions of the main components of SOA modeling. Its extension on a specific domain can help to construct a knowledge base for business process modeling, describing processes and services and defining their mappings. As such, we propose four modeling methods for SOA development: TDM (Top-Down Modeling) supports developers to directly create new process models; TDM-RP (Top-Down Modeling based on Reusable Process) supports developers to construct new process models by reusing similar process models already defined in the knowledge base; BUM-RS (Bottom-Up Modeling based on Reusable Services) supports developers to construct new process models by reusing identified services; and AM-RPRS (Agile Modeling based on Reusable Process and Reusable Services) provides specific queries on both the business process and identified services, and enables developers to efficiently choose suitable models. The modeling methods are verified by extending BPO to the automotive software development domain, constructing a knowledge base AutoPO (Automotive Process Ontology), and applying AutoPO to simulate the execution of the modeling method with a series of case studies. Based on a survey of the quality requirements for models, we also propose a set of quality attributes for SOA models.