SPECTRAL BASED METHODS THAT STREAMLINE THE SEARCH FOR FAILURE SCENARIOS IN LARGE-SCALE DISTRIBUTED SYSTEMS

摘要

We report our work on the development of analytical andrnnumerical methods that enable the detection of failure scenariosrnin distributed grid computing, cloud computing andrnother large scale systems.The spectral (i.e. eigenvalue andrneigenvector) properties of the matrices associated with arnnon-homogeneous absorbing Markov Chain are used tornquickly compute the long time proportion of tasks completedrnat a given setting of parameters. This enables therndiscovery of critical ranges of parameter values where systemrnperformance deteriorates and fails.