Concepts for openMP target offload resilience

Christian Engelmann; Geoffroy R. Vallée; Swaroop Pophale

doi:10.1007/978-3-030-28596-8_6

Concepts for openMP target offload resilience

Christian Engelmann, Geoffroy R. Vallée, Swaroop Pophale

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

1 Scopus citations

Abstract

Recent reliability issues with one of the fastest supercomputers in the world, Titan at Oak Ridge National Laboratory (ORNL), demonstrated the need for resilience in large-scale heterogeneous computing. OpenMP currently does not address error and failure behavior. This paper takes a first step toward resilience for heterogeneous systems by providing the concepts for resilient OpenMP offload to devices. Using real-world error and failure observations, the paper describes the concepts and terminology for resilient OpenMP target offload, including error and failure classes and resilience strategies. It details the experienced general-purpose computing graphics processing unit (GPGPU) errors and failures in Titan. It further proposes improvements in OpenMP, including a preliminary prototype design, to support resilient offload to devices for efficient handling of errors and failures in heterogeneous high-performance computing (HPC) systems.

Original language	English
Title of host publication	OpenMP
Subtitle of host publication	Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings
Editors	Xing Fan, Oliver Sinnen, Nasser Giacaman, Bronis R. de Supinski
Publisher	Springer Verlag
Pages	78-93
Number of pages	16
ISBN (Print)	9783030285951
DOIs	https://doi.org/10.1007/978-3-030-28596-8_6
State	Published - 2019
Event	15th International Workshop on OpenMP, IWOMP 2019 - Auckland, New Zealand Duration: Sep 11 2019 → Sep 13 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11718 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Workshop on OpenMP, IWOMP 2019
Country/Territory	New Zealand
City	Auckland
Period	09/11/19 → 09/13/19

Funding

Research sponsored by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. Department of Energy. This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paidup, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/ downloads/doe-public-access-plan).

Keywords

OpenMP
Resilience
Supercomputing

Access to Document

10.1007/978-3-030-28596-8_6

Cite this

Engelmann, C., Vallée, G. R., & Pophale, S. (2019). Concepts for openMP target offload resilience. In X. Fan, O. Sinnen, N. Giacaman, & B. R. de Supinski (Eds.), OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings (pp. 78-93). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11718 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-28596-8_6

Engelmann, Christian ; Vallée, Geoffroy R. ; Pophale, Swaroop. / Concepts for openMP target offload resilience. OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. editor / Xing Fan ; Oliver Sinnen ; Nasser Giacaman ; Bronis R. de Supinski. Springer Verlag, 2019. pp. 78-93 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{643359a60eae4309af9ebe3d8d970a87,

title = "Concepts for openMP target offload resilience",

abstract = "Recent reliability issues with one of the fastest supercomputers in the world, Titan at Oak Ridge National Laboratory (ORNL), demonstrated the need for resilience in large-scale heterogeneous computing. OpenMP currently does not address error and failure behavior. This paper takes a first step toward resilience for heterogeneous systems by providing the concepts for resilient OpenMP offload to devices. Using real-world error and failure observations, the paper describes the concepts and terminology for resilient OpenMP target offload, including error and failure classes and resilience strategies. It details the experienced general-purpose computing graphics processing unit (GPGPU) errors and failures in Titan. It further proposes improvements in OpenMP, including a preliminary prototype design, to support resilient offload to devices for efficient handling of errors and failures in heterogeneous high-performance computing (HPC) systems.",

keywords = "OpenMP, Resilience, Supercomputing",

author = "Christian Engelmann and Vall{\'e}e, {Geoffroy R.} and Swaroop Pophale",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 15th International Workshop on OpenMP, IWOMP 2019 ; Conference date: 11-09-2019 Through 13-09-2019",

year = "2019",

doi = "10.1007/978-3-030-28596-8_6",

language = "English",

isbn = "9783030285951",

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

publisher = "Springer Verlag",

pages = "78--93",

editor = "Xing Fan and Oliver Sinnen and Nasser Giacaman and {de Supinski}, {Bronis R.}",

booktitle = "OpenMP",

}

Engelmann, C, Vallée, GR & Pophale, S 2019, Concepts for openMP target offload resilience. in X Fan, O Sinnen, N Giacaman & BR de Supinski (eds), OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11718 LNCS, Springer Verlag, pp. 78-93, 15th International Workshop on OpenMP, IWOMP 2019, Auckland, New Zealand, 09/11/19. https://doi.org/10.1007/978-3-030-28596-8_6

Concepts for openMP target offload resilience. / Engelmann, Christian; Vallée, Geoffroy R.; Pophale, Swaroop.
OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. ed. / Xing Fan; Oliver Sinnen; Nasser Giacaman; Bronis R. de Supinski. Springer Verlag, 2019. p. 78-93 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11718 LNCS).

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

TY - GEN

T1 - Concepts for openMP target offload resilience

AU - Engelmann, Christian

AU - Vallée, Geoffroy R.

AU - Pophale, Swaroop

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - Recent reliability issues with one of the fastest supercomputers in the world, Titan at Oak Ridge National Laboratory (ORNL), demonstrated the need for resilience in large-scale heterogeneous computing. OpenMP currently does not address error and failure behavior. This paper takes a first step toward resilience for heterogeneous systems by providing the concepts for resilient OpenMP offload to devices. Using real-world error and failure observations, the paper describes the concepts and terminology for resilient OpenMP target offload, including error and failure classes and resilience strategies. It details the experienced general-purpose computing graphics processing unit (GPGPU) errors and failures in Titan. It further proposes improvements in OpenMP, including a preliminary prototype design, to support resilient offload to devices for efficient handling of errors and failures in heterogeneous high-performance computing (HPC) systems.

AB - Recent reliability issues with one of the fastest supercomputers in the world, Titan at Oak Ridge National Laboratory (ORNL), demonstrated the need for resilience in large-scale heterogeneous computing. OpenMP currently does not address error and failure behavior. This paper takes a first step toward resilience for heterogeneous systems by providing the concepts for resilient OpenMP offload to devices. Using real-world error and failure observations, the paper describes the concepts and terminology for resilient OpenMP target offload, including error and failure classes and resilience strategies. It details the experienced general-purpose computing graphics processing unit (GPGPU) errors and failures in Titan. It further proposes improvements in OpenMP, including a preliminary prototype design, to support resilient offload to devices for efficient handling of errors and failures in heterogeneous high-performance computing (HPC) systems.

KW - OpenMP

KW - Resilience

KW - Supercomputing

UR - http://www.scopus.com/inward/record.url?scp=85072862934&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-28596-8_6

DO - 10.1007/978-3-030-28596-8_6

M3 - Conference contribution

AN - SCOPUS:85072862934

SN - 9783030285951

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

SP - 78

EP - 93

BT - OpenMP

A2 - Fan, Xing

A2 - Sinnen, Oliver

A2 - Giacaman, Nasser

A2 - de Supinski, Bronis R.

PB - Springer Verlag

T2 - 15th International Workshop on OpenMP, IWOMP 2019

Y2 - 11 September 2019 through 13 September 2019

ER -

Engelmann C, Vallée GR, Pophale S. Concepts for openMP target offload resilience. In Fan X, Sinnen O, Giacaman N, de Supinski BR, editors, OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. Springer Verlag. 2019. p. 78-93. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-28596-8_6

Concepts for openMP target offload resilience

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this