Online Failure Prediction of Dynamically Evolving Systems

When dealing with dynamically evolving systems, components may be inserted or removed while the system is being operated. Unsafe run-time changes may compromise the correct execution of the entire system, and online analysis techniques have been proposed to mitigate the effects of such failures. With CASSANDRA we want to move one step ahead, by defining a novel proactive monitoring and verification technique with the ability to predict (and prevent) the potential failures happening in the future. CASSANDRA combines designtime and run-time information for implementing an online failure prediction approach: run-time information is used to identify the current execution state, and to drive the construction of a design-time model that looks k steps ahead of the current execution state. Such a model is used to check whether a set of desired properties will be potentially neglected in near future executions. CASSANDRA algorithms have been formally defined and the approach has been concretized into the OSGi Platform, applied to a realistic case study and evaluated.