H Hamidi - Islamic Azad University -Doroud Branch

We present a new approach to algorithm based fault tolerance (ABFT) for High Performance Computing system. The Algorithm Based Fault Tolerance approach transforms a system that does not tolerate a specific type of faults, called the fault-intolerant system, to a system that provides a specific level of fault tolerance, namely recovery. We have implemented a systematic procedure for introducing structured redundancy into ABFT. Algorithm Based Fault Tolerance has been recommending as a cost-effective concurrent error detection scheme. It proposes a novel computing paradigm to provide fault tolerance for numerical algorithms. To that end, a matrix-based model has been developed and, based on that, algorithms for both the design and analysis of ABFT systems are formulated

Algorithm Based Fault Tolerance (ABFT), Checkpointing, Error Correction, Matrix operations