Distributed, dynamic, and efficient testing of large scale multiple processor systems

The authors propose a distributed test algorithm for multiple processor systems where a central fault-free entity such as a host computer does not exist for any purpose such as control, fault-diagnosis, and system reconfiguration. The proposed algorithm is dynamic, i.e. testing assignments are not fixed ahead of time but rather are made as processors or interprocessor links fail. This dynamic property of the algorithm improves system performance by reducing the number of testers assigned to every processor while allowing the existence of multiple faulty processors and interprocessor communication links at every round of testing. Simulation results verify the effectiveness of the given approach