Skip to Main content Skip to Navigation
Reports

Combining algorithm-based fault tolerance and checkpointing for iterative solvers

Abstract : Several recent papers have introduced a periodic verification mechanism to detect silent errors in iterative solvers. Chen [PPoPP'13, pp. 167--176] has shown how to combine such a verification mechanism (a stability test checking the orthogonality of two vectors and recomputing the residual) with checkpointing: the idea is to verify every $d$ iterations, and to checkpoint every $c \times d$ iterations. When a silent error is detected by the verification mechanism, one can rollback to and re-execute from the last checkpoint. In this paper, we also propose to combine checkpointing and verification, but we use ABFT rather than stability tests. ABFT can be used for error detection, but also for error detection and correction, allowing a forward recovery (and no rollback nor re-execution) when a single error is detected. We present a performance model to compute the performance of all schemes, and we instantiate it using the Conjugate Gradient algorithm. Finally, we validate our new approach through a set of simulations.
Complete list of metadata

Cited literature [33 references]  Display  Hide  Download

https://hal.inria.fr/hal-01111707
Contributor : Equipe Roma <>
Submitted on : Tuesday, February 24, 2015 - 8:41:12 PM
Last modification on : Monday, December 14, 2020 - 3:32:30 PM
Long-term archiving on: : Sunday, April 16, 2017 - 11:10:15 AM

File

RR-8675.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01111707, version 2

Collections

Citation

Massimiliano Fasi, Yves Robert, Bora Uçar. Combining algorithm-based fault tolerance and checkpointing for iterative solvers. [Research Report] RR-8675, INRIA Grenoble - Rhône-Alpes; INRIA. 2015. ⟨hal-01111707v2⟩

Share

Metrics

Record views

498

Files downloads

645