Fault tolerance in an inner-outer solver: A GVR-enabled case study

Ziming Zheng; Andrew A. Chien; Keita Teranishi

doi:10.1007/978-3-319-17353-5_11

Fault tolerance in an inner-outer solver: A GVR-enabled case study

Ziming Zheng, Andrew A. Chien, Keita Teranishi

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Scopus citations

Abstract

Resilience is a major challenge for large-scale systems. It is particularly important for iterative linear solvers, since they take much of the time of many scientific applications. We show that single bit flip errors in the Flexible GMRES iterative linear solver can lead to high computational overhead or even failure to converge to the right answer. Informed by these results, we design and evaluate several strategies for fault tolerance in both inner and outer solvers appropriate across a range of error rates.We implement them, extending Trilinos’ solver library with the Global View Resilience (GVR) programming model, which provides multi-stream snapshots, multi-version data structures with portable and rich error checking/recovery. Experimental results validate correct execution with low performance overhead under varied error conditions.

Original language	English
Title of host publication	High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers
Editors	Osni Marques, Michel Dayde, Kengo Nakajima
Publisher	Springer Verlag
Pages	124-132
Number of pages	9
ISBN (Print)	9783319173528
DOIs	https://doi.org/10.1007/978-3-319-17353-5_11
State	Published - 2015
Externally published	Yes
Event	11th International Conference on High Performance Computing for Computational Science, VECPAR 2014 - Eugene, United States Duration: Jun 30 2014 → Jul 3 2014

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	8969
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th International Conference on High Performance Computing for Computational Science, VECPAR 2014
Country/Territory	United States
City	Eugene
Period	06/30/14 → 07/3/14

Funding

We thank Mark Hoemmen from Sandia National Laboratories for his advice. This work supported by the Office of Advanced Scientific Computing Research, Office of Science, U.S. Department of Energy, under Award DE-SC0008603 and Contract DE-AC02-06CH11357. Also under the DOE National Nuclear Security Administration (NNSA) Advanced Simulation and Computing (ASC) program. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy National Nuclear Security Administration under contract DE-AC04-94AL85000.

Keywords

High performance computing
Numerical solver
Resilience

Access to Document

10.1007/978-3-319-17353-5_11

Cite this

Zheng, Z., Chien, A. A., & Teranishi, K. (2015). Fault tolerance in an inner-outer solver: A GVR-enabled case study. In O. Marques, M. Dayde, & K. Nakajima (Eds.), High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers (pp. 124-132). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8969). Springer Verlag. https://doi.org/10.1007/978-3-319-17353-5_11

Zheng, Ziming ; Chien, Andrew A. ; Teranishi, Keita. / Fault tolerance in an inner-outer solver : A GVR-enabled case study. High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. editor / Osni Marques ; Michel Dayde ; Kengo Nakajima. Springer Verlag, 2015. pp. 124-132 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{32bb5d898f564837980dad024cb48bdd,

title = "Fault tolerance in an inner-outer solver: A GVR-enabled case study",

abstract = "Resilience is a major challenge for large-scale systems. It is particularly important for iterative linear solvers, since they take much of the time of many scientific applications. We show that single bit flip errors in the Flexible GMRES iterative linear solver can lead to high computational overhead or even failure to converge to the right answer. Informed by these results, we design and evaluate several strategies for fault tolerance in both inner and outer solvers appropriate across a range of error rates.We implement them, extending Trilinos{\textquoteright} solver library with the Global View Resilience (GVR) programming model, which provides multi-stream snapshots, multi-version data structures with portable and rich error checking/recovery. Experimental results validate correct execution with low performance overhead under varied error conditions.",

keywords = "High performance computing, Numerical solver, Resilience",

author = "Ziming Zheng and Chien, \{Andrew A.\} and Keita Teranishi",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014 ; Conference date: 30-06-2014 Through 03-07-2014",

year = "2015",

doi = "10.1007/978-3-319-17353-5\_11",

language = "English",

isbn = "9783319173528",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "124--132",

editor = "Osni Marques and Michel Dayde and Kengo Nakajima",

booktitle = "High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers",

}

Zheng, Z, Chien, AA & Teranishi, K 2015, Fault tolerance in an inner-outer solver: A GVR-enabled case study. in O Marques, M Dayde & K Nakajima (eds), High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8969, Springer Verlag, pp. 124-132, 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014, Eugene, United States, 06/30/14. https://doi.org/10.1007/978-3-319-17353-5_11

Fault tolerance in an inner-outer solver: A GVR-enabled case study. / Zheng, Ziming; Chien, Andrew A.; Teranishi, Keita.
High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. ed. / Osni Marques; Michel Dayde; Kengo Nakajima. Springer Verlag, 2015. p. 124-132 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8969).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Fault tolerance in an inner-outer solver

T2 - 11th International Conference on High Performance Computing for Computational Science, VECPAR 2014

AU - Zheng, Ziming

AU - Chien, Andrew A.

AU - Teranishi, Keita

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2015.

PY - 2015

Y1 - 2015

N2 - Resilience is a major challenge for large-scale systems. It is particularly important for iterative linear solvers, since they take much of the time of many scientific applications. We show that single bit flip errors in the Flexible GMRES iterative linear solver can lead to high computational overhead or even failure to converge to the right answer. Informed by these results, we design and evaluate several strategies for fault tolerance in both inner and outer solvers appropriate across a range of error rates.We implement them, extending Trilinos’ solver library with the Global View Resilience (GVR) programming model, which provides multi-stream snapshots, multi-version data structures with portable and rich error checking/recovery. Experimental results validate correct execution with low performance overhead under varied error conditions.

AB - Resilience is a major challenge for large-scale systems. It is particularly important for iterative linear solvers, since they take much of the time of many scientific applications. We show that single bit flip errors in the Flexible GMRES iterative linear solver can lead to high computational overhead or even failure to converge to the right answer. Informed by these results, we design and evaluate several strategies for fault tolerance in both inner and outer solvers appropriate across a range of error rates.We implement them, extending Trilinos’ solver library with the Global View Resilience (GVR) programming model, which provides multi-stream snapshots, multi-version data structures with portable and rich error checking/recovery. Experimental results validate correct execution with low performance overhead under varied error conditions.

KW - High performance computing

KW - Numerical solver

KW - Resilience

UR - https://www.scopus.com/pages/publications/84942594970

U2 - 10.1007/978-3-319-17353-5_11

DO - 10.1007/978-3-319-17353-5_11

M3 - Conference contribution

AN - SCOPUS:84942594970

SN - 9783319173528

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 124

EP - 132

BT - High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers

A2 - Marques, Osni

A2 - Dayde, Michel

A2 - Nakajima, Kengo

PB - Springer Verlag

Y2 - 30 June 2014 through 3 July 2014

ER -

Zheng Z, Chien AA, Teranishi K. Fault tolerance in an inner-outer solver: A GVR-enabled case study. In Marques O, Dayde M, Nakajima K, editors, High Performance Computing for Computational Science - VECPAR 2014 - 11th International Conference, Revised Selected Papers. Springer Verlag. 2015. p. 124-132. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-17353-5_11

Fault tolerance in an inner-outer solver: A GVR-enabled case study

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this