Enabling Proactive Adaptation through Just-in-time Testing of Conversational Services

Service-based applications (SBAs) will increasingly be composed of third-party services available over the Internet. Reacting to failures of those third-party services by dynamically adapting the SBAs will become a key enabler for ensuring reliability. Determining when to adapt an SBA is especially challenging in the presence of conversational (aka. stateful) services. A conversational service might fail in the middle of an invocation sequence, in which case adapting the SBA might be costly; e.g., due to the necessary state transfer to an alternative service. In this paper we propose just-in-time testing of conversational services as a novel approach to detect potential problems and to proactively trigger adaptations, thereby preventing costly compensation activities. The approach is based on a framework for online testing and a formal test-generation method which guarantees functional correctness for conversational services. The applicability of the approach is discussed with respect to its underlying assumptions and its performance. The benefits of the approach are demonstrated using a realistic example.