On-line fault classification and handling in IEEE1687 based fault management system for complex SoCs

Semiconductor products manufactured with latest and emerging processes are increasingly prone to wear out and aging. While the fault occurrence rate in such systems increases, the fault tolerance techniques are becoming even more expensive and one cannot rely on them alone. In addition to mitigating/correcting the faults, the system may systematically monitor, detect, localize, diagnose and classify them (manage faults). As a result of such fault management approach, the system may continue operating and degrade gracefully even in case if some of the system's resources become unusable due to intolerable faults. This works proposes a fault classification and handling methodology that fits to an event-driven on-line fault monitoring, signaling and management architecture based on IEEE1687 IJTAG and suitable for a modern complex SoC with many heterogeneous cores.