Automated Prevention of Failure in Complex and Large Systems: Fighting Fire with Fire

People, businesses and economies are increasingly dependent on Cloud and internet services. At the same time, the systems on which these services are built are becoming more complex and interdependent. The cost of failure is high, but systems are too complex for human detection of problems. We review methods for online fault diagnosis, process mining and Virtual Machine Introspection. We suggest bringing these techniques together for automated identification, diagnosis and prediction of risk of failure in large systems. We present examples from telecoms and Cloud industries in support of these ideas.