Chronicles for On-line Diagnosis of Distributed Systems

The formalism of chronicles has been proposed a few years ago to monitor and diagnose dynamic physical systems and has been successfully used for real-time appli- cations. A chronicle, expressed by a set of time-constrained events, describes the situations to monitor. Even if efficient algorithms exist to analyze a flow of observed events and to rec- ognize on-the-fly the matching chronicles, it is now well-known that distributed approaches are better suited to monitor actual systems. In this report, we adapt the chronicle-based approach to a distributed context. We propose a decentralized architecture and the cor- responding software, implemented under the name of CarDeCRS (Chronicles Applied to error Recognition in Distributed Environments, through CRS), which in charge of synchro- nizing the local diagnoses, computed by chronicle-based local diagnosers, and merging them into a global diagnosis. This work is motivated by an application that aims at monitoring the behaviour of software components within the WS-Diamond European project. In this context, a request is sent to a web service which collaborates with other services to provide the adequate reply. Faults may propagate from one service to another and diagnosing them is a crucial issue, in order to react properly. We use a simplified example of a e-foodshop web service to illustrate our proposal