This paper presents and evaluates a hybrid fault tolerance approach for dynamically scheduled processors that combines on-line error-correction for run-time fault handling with reconfiguration techniques for permanent fault handling. A permanent reconfiguration is triggered on-demand during run-time, depending on the frequency of on-line corrected faults. The presented work evaluates the effect of the permanent reconfiguration regarding the reliability and performance impact. It turns out that only in few cases the on-line handling of permanent faults performs better than the reconfiguration-mechanism.
展开▼