On-line failure prediction in safety-critical systems

The paper proposes a novel architecture, namely CASPER, for online failure prediction that has the distinctive features to be (i) black-box: no knowledge of applications internals and logic of the system is required (ii) non-intrusive: no status information of the components is used such as CPU or memory usage; The architecture has been implemented to predict failures in a real Air Traffic Control System. CASPER exhibits high degree of accuracy in predicting failures with low false positive rate. The experimental validation shows how operators are provided with predictions issued a few hundred of seconds before the occurrence of the failure.