Online black-box failure prediction for mission critical distributed systems

This paper introduces a novel approach to failure prediction for mission critical distributed systems that has the distinctive features to be black-box, non-intrusive and online. The approach combines Complex Event Processing (CEP) and Hidden Markov Models (HMM) so as to analyze symptoms of failures that might occur in the form of anomalous conditions of performance metrics identified for such purpose. The paper describes an architecture named CASPER, based on CEP and HMM, that relies on sniffed information from the communication network of a mission critical system, only, for predicting anomalies that can lead to software failures. An instance of CASPER has been implemented, trained and tuned to monitor a real Air Traffic Control (ATC) system. An extensive experimental evaluation of CASPER is presented. The obtained results show (i) a very low percentage of false positives over both normal and under stress conditions, and (ii) a sufficiently high failure prediction time that allows the system to apply appropriate recovery procedures. © 2012 Springer-Verlag