End-to-end fault containment in scalable shared-memory multiprocessors

Current shared-memory multiprocessors suffer from an inherent fragility, since a single hardwareor system software failure can cause the entire machine to crash. This dissertation describes acombination of hardware and software techniques that can be used to provide fault containmentfor large-scale shared memory machines. With fault containment, the impact of a fault remainslimited to only a small portion of the system, while the remaining good parts can continueoperating normally after...