SOAR: An Extended Model-Based Reasoning for Diagnosing Faults in Service-Oriented Architecture

Service-oriented architecture (SOA) is a cost effective approach to building enterprise applications. SOA reveals non-conventional characteristics of heterogeneity, grid-like distribution, evolvability, and limited visibility. Hence, services management presents non-conventional challenges. Especially, fault diagnosis at runtime is challenging due to the SOA features. Model-based reasoning (MBR) is a formal approach to diagnosing faults, which is based on predicate calculus and term resolution. In this paper, we present SOAR (service-oriented abductive reasoning) which extends the basic MBR to diagnose faults in various SOA components. SOAR provides an enhanced inference capability with state-based and QoS-based reasoning in addition to the basic setting/observation-based reasoning. We propose concrete schemes to formally represent system description, normal behavior, fault model and observations, and reasoning methods to diagnose faults and to determine their causes. In addition, we present a case study of applying SOAR to show how it is applied in practice and how the diagnosis can be conducted in autonomous way.