A reconfigurable testing approach for service-based applications

Internet-based software applications, including applications operated by third parties, can be progressively integrated on a dynamic basis to provide useful functionality. Such dynamic behaviour is made possible through the use of Service Oriented Architecture (SOA). While SOA brings substantial advantages to software development and deployment, it is technically inefficient to test these software applications with existing software testing techniques. Based on analysis of service-based applications; it is evident that validating such applications requires an automated test process that performs testing when the applications are running and deployed in a 'live' environment. The literature revealed limitations such as adaptability of test systems, automated test data generation and maintaining test data integrity when services change, in testing mechanisms available. It also provided promising research directions for addressing the problem of adaptability in testing systems during runtime testing of service-based applications. This thesis presents an automated testing solution that provides adaptive runtime testing for Web services. The solution was implemented in two phases; firstly, a preliminary study was conducted to analyse the feasibility of continuing with the development of a testing system. The preliminary study resulted in implementation of the Alpha Test System (ATS). The ATS approach was extended in a detailed study that implemented a second system, called the Beta Test System (BTS). The proofs-ofconcept systems performed runtime testing on atomic as well as composite servicebased applications, conducting runtime testing and providing live feed back to the service providers. The testing systems were also able to adapt their behaviour to support changes in the service interface level by using a new method presented in this thesis. The BTS incorporated a novel mechanism to identify internal composition changes in a composite Web service while it is live. Evaluation of the proofs-of-concept revealed that the new approach presented in this thesis addressed the limitation of providing adaptive test system for testing servicebased applications. This approach can be further developed for commercial implementations. The evaluation also revealed that the automated testing systems require a combination of syntactical and semantic information in the service specifications which would leverage test system adaptation and test data generation. In conclusion, the proofs-of-concept could be used to test applications that are deployed in other service-based computing paradigms; such as, 'Cloud Computing'.