Spectrum-based fault diagnosis for service-oriented software systems

Due to the loosely coupled and highly dynamic nature of service-oriented systems, the actual configuration of such system only fully materializes at runtime, rendering many of the traditional quality assurance approaches useless. In order to enable service systems to recover from and adapt to runtime failures, an important step is to detect failures and diagnose problematic services automatically. This paper presents a lightweight, fully automated, spectrum-based diagnosis technique for service-oriented software systems that is combined with a framework-based online monitor. An experiment with a case system is set up to validate the feasibility of pinpointing problematic service operations. The results indicate that this approach is able to identify problematic service operations correctly in 72% of the cases.