Dynamic Selective Protection of Sparse Iterative Solvers via ML Prediction of Soft Error Impacts.

Soft errors occur frequently on large computing platforms due to the increasing scale and complexity of HPC systems. Various resilience techniques (e.g., checkpointing, ABFT, and replication) have been proposed to protect scientific applications from soft errors at different levels. Among them, system-level replication often involves duplicating or even triplicating the entire computation, thus resulting in high resilience overhead. This paper proposes dynamic selective protection for sparse iterative solvers, in particular for the Preconditioned Conjugate Gradient (PCG) solver, at the system level to reduce the resilience overhead. For this method, we leverage machine learning (ML) to predict the impact of soft errors that strike different elements of a key computation (i.e., sparse matrix-vector multiplication) at different iterations of the solver. Based on the result of the prediction, we design a dynamic strategy to selectively protect those elements that would result in a large performance degradation if struck by soft errors. An experimental evaluation demonstrates that our dynamic protection strategy is able to reduce the resilience overhead compared to existing algorithms.