Augmented Diagnosis and Testing for SOAs - Audit 4 SOAs

The principles of Service-Oriented Architectures (SOAs) have been receiving high attention lately, and are nowadays widely established in the software industry for the application in many domains. The fast evolution of available technology entails the development of new complex domains, and in turn enables and drives active research in fields such as self-optimizing distributed systems, autonomic computing, adaptive systems, or cloud computing. Nowadays SOA-based systems consist of a multitude of Web services, registries/brokers, mediators, message buses, monitors, governance systems and many more components. This ever-growing complexity leads however to a rather high error-proneness, due to the large amount of components and complex as well as subtle links and dependencies between individual system parts. Today there exist numerous solutions that assist designers in testing single Web services. We are however faced with a lack of options that enable methodical system-wide runtime tests of complex SOA systems. We argue that a system-wide high-quality assessment in real conditions is an essential factor in the process of facing today`s complexity of SOAs and assure the quality of related prototypes and production systems while keeping development time within manageable bounds. With this project, we face the challenge to provide a methodology and related technologies for a high-quality development environment for SOAs. Specifically we will facilitate the development of SOA models that enable an automated construction of realistic testbeds and high-quality test-suites for system-wide runtime tests. These runtime tests will enable the assessment of important system traits like performance, stability, robustness, scalability, and other functional and non-functional properties in real world scenarios and conditions. Furthermore, we will integrate diagnostic reasoning in the model itself and the testing process. Developed models will thus also contain diagnostic reasoning aspects, so that for failing tests we will provide engineers and researchers with the means to isolate faults using model-based diagnosis techniques. The integration of ideas from the testing as well as model-based diagnosis communities in an ``active diagnosis`` step will enable us to derive additional tests for the discrimination between ambiguous diagnoses, an essential factor in our proposed work-flow. The diagnostic aspects of the model will also act as additional and interesting criteria for the test-suite generation process. With our focus on providing a methodology and related technologies for a high-quality assessment of SOA traits under realistic conditions, we address current needs of researchers and engineers in this area. We will enable them to assure the quality of their prototypes and products by finding errors with high quality system-wide runtime tests and the option to isolate corresponding faults using diagnostic reasoning. The models will also allow very early design assessment. In the course of the extensive research necessary to achieve our aims, we will integrate the know-how from multiple communities (SOA, testing, model-based diagnosis). As a consequence, in our vision, we expect to give these communities new impulses by the know-how transfer resulting from this project, which, besides our achievements, will sparkle additional research by other groups.

Service-oriented architectures (SOAs) have become an attractive means to implement a softwares functionality by flexibly drawing on individual web-services that offer an easily accessible network-oriented interface to ones own or a third partys intellectual property. Via their flexibility and scalability, they allow us to design also large-scale products in the cloud (an implementation of SOA) that can adapt to a customers demands. Ensuring that designed software lives up to the expected quality is very challenging, since in addition to typical software development issues, domain specific issues like controllability of external web services or limited knowledge about them have to be addressed, as well as the fact that SOAs can offer concurrent computations. In this context, Audit4SOAs has been focusing on providing the theoretical foundations for an integrated testing and diagnosis workflow that supports engineers in the design and verification of corresponding SOA designs, as well as maintain reliable and dependable operation at runtime. Specific research questions that we tackled in our research were the question of how to effectively model SOAs and how to efficiently derive test suites that are effective in unveiling faults in the SOA under scrutiny. To this end, we captured a SOAs structure in a control flow graph that can be annotated with further knowledge about the system or individual parts. Using a constraint representation of this graph, we showed how to efficiently derive effective test suites, exploring several strategies implementing structural and random aspects in the search for test cases. For a failed test case, there is the issue and obvious task of localizing the fault(s) responsible for the unexpected behavior. To this end, we developed a model-based diagnosis/debugging approach that focuses on such a failed execution and offers diagnoses in the form of component subsets that can explain the issue if we assume those components (program statements) to be faulty. With active diagnosis techniques, we can furthermore design new tests automatically, that provide further data to discriminate between diagnoses (if this is possible) in that this data renders some of the diagnosis candidates unviable. In addition to automated test case generation, we developed techniques for fault identification, automated prediction of SLA violations, and optimizing QoS parameters of complex SOAs at runtime. We furthermore explored novel mechanisms for adaptive resource allocation and controlled, testable evolution of dependable, service-based systems.