A systematic fault-tolerant computational model for both crash failures and silent data corruption

Abstract

As the boundaries between Cloud and HPC continue to blur, it is clear that there is an urgent demand for a systematic computational model that adapts to the computing platform and accommodates the underlying workloads. As computing systems continue to scale out to satisfy the increasingly large demands on computing capacity, power awareness and fault tolerance have become major concerns. This paper proposes a novel computational model that applies to both compute- and data-intensive workloads, and deals with diverse types of faults. Evaluation results demonstrate that the proposed model is able to achieve significant energy savings compared to existing fault tolerance techniques, while maintaining the same level of fault tolerance.