No abstract

Service Oriented Architecture (SOA) is a common design pattern that allows building applications composed of several services. It promotes features as interoperability, scalability, and software reuse. Services composing a SOA system may evolve and change during runtime, often outside the control of the owner of the application, which makes the verification and validation processes complex. Among all the automated techniques to validate the behavior of an SOA application, is Model-Based Testing (MBT). MBT requires an accurate model of the application in order to generate suitable test cases. However, the intrinsic of a SOA application sets significant challenges to MBT effectiveness. In this paper we discuss the challenges in the testing of SOA applications, and we propose the use of Model-Driven Engineering (MDE) to improve the flexibility of testing tools. Finally, we outline our plan for realizing MDE-driven MBT within an existing online testing framework.

Online model-based testing is one of the most suitable techniques to assess the proper behavior of service orchestrations. However, the diverse panorama in terms of modeling languages and test case generation tools is a limitation to widespread adoption. We advocate that the application of Model-Driven Engineering principles as meta-modeling and model transformation can cope with this problem, improving the interoperability of artifacts in the test case generation process, thus bringing benefits in case of agile development processes, where system and technology evolution is frequent. In this paper, we present our contribution to this idea, introducing i) a reference metamodel, which stores the business process behavior and the information to generate input models for testing tools, and ii) transformations from orchestration languages towards testing tools. The proposed approach is implemented in a testing framework and evaluated on a case study where multiple orchestrations are expressed in two languages. Also, the paper presents how test cases are appropriately generated and successfully executed, starting from an orchestration model as a consequence of successful transformations.