Scalability of partial differential equations preconditioner resilient to soft and hard faults

Karla Morris; Francesco Rizzi; Khachik Sargsyan; Kathryn Dahlgren; Paul Mycek; Cosmin Safta; Olivier Le Maître; Omar Knio; Bert Debusschere

doi:10.1007/978-3-319-41321-1_24

Scalability of partial differential equations preconditioner resilient to soft and hard faults

Karla Morris^*, Francesco Rizzi, Khachik Sargsyan, Kathryn Dahlgren, Paul Mycek, Cosmin Safta, Olivier Le Maître, Omar Knio, Bert Debusschere

^*Corresponding author for this work

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

Abstract

We present a resilient domain-decomposition preconditioner for partial differential equations (PDEs). The algorithm reformulates the PDE as a sampling problem, followed by a solution update through data manipulation that is resilient to both soft and hard faults. We discuss an implementation based on a server-client model where all state information is held by the servers, while clients are designed solely as computational units. Servers are assumed to be “sandboxed”, while no assumption is made on the reliability of the clients. We explore the scalability of the algorithm up to ∼12k cores, build an SST/macro skeleton to extrapolate to∼50k cores, and show the resilience under simulated hard and soft faults for a 2D linear Poisson equation.

Original language	English (US)
Title of host publication	High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings
Editors	Jack Dongarra, Julian M. Kunkel, Pavan Balaji
Publisher	Springer Verlag
Pages	469-485
Number of pages	17
ISBN (Print)	9783319413204
DOIs	https://doi.org/10.1007/978-3-319-41321-1_24
State	Published - 2016
Event	31st International Conference on High Performance Computing, ISC High Performance 2016 - Frankfurt, Germany Duration: Jun 19 2016 → Jun 23 2016

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9697
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	31st International Conference on High Performance Computing, ISC High Performance 2016
Country/Territory	Germany
City	Frankfurt
Period	06/19/16 → 06/23/16

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-319-41321-1_24

Cite this

Morris, K., Rizzi, F., Sargsyan, K., Dahlgren, K., Mycek, P., Safta, C., Le Maître, O., Knio, O., & Debusschere, B. (2016). Scalability of partial differential equations preconditioner resilient to soft and hard faults. In J. Dongarra, J. M. Kunkel, & P. Balaji (Eds.), High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings (pp. 469-485). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9697). Springer Verlag. https://doi.org/10.1007/978-3-319-41321-1_24

Morris, Karla ; Rizzi, Francesco ; Sargsyan, Khachik et al. / Scalability of partial differential equations preconditioner resilient to soft and hard faults. High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings. editor / Jack Dongarra ; Julian M. Kunkel ; Pavan Balaji. Springer Verlag, 2016. pp. 469-485 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{d0640ce1d74f48949423c93d942ef015,

title = "Scalability of partial differential equations preconditioner resilient to soft and hard faults",

abstract = "We present a resilient domain-decomposition preconditioner for partial differential equations (PDEs). The algorithm reformulates the PDE as a sampling problem, followed by a solution update through data manipulation that is resilient to both soft and hard faults. We discuss an implementation based on a server-client model where all state information is held by the servers, while clients are designed solely as computational units. Servers are assumed to be “sandboxed”, while no assumption is made on the reliability of the clients. We explore the scalability of the algorithm up to ∼12k cores, build an SST/macro skeleton to extrapolate to∼50k cores, and show the resilience under simulated hard and soft faults for a 2D linear Poisson equation.",

author = "Karla Morris and Francesco Rizzi and Khachik Sargsyan and Kathryn Dahlgren and Paul Mycek and Cosmin Safta and {Le Ma{\^i}tre}, Olivier and Omar Knio and Bert Debusschere",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2016.; 31st International Conference on High Performance Computing, ISC High Performance 2016 ; Conference date: 19-06-2016 Through 23-06-2016",

year = "2016",

doi = "10.1007/978-3-319-41321-1_24",

language = "English (US)",

isbn = "9783319413204",

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

publisher = "Springer Verlag",

pages = "469--485",

editor = "Jack Dongarra and Kunkel, {Julian M.} and Pavan Balaji",

booktitle = "High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings",

address = "Germany",

}

Morris, K, Rizzi, F, Sargsyan, K, Dahlgren, K, Mycek, P, Safta, C, Le Maître, O, Knio, O & Debusschere, B 2016, Scalability of partial differential equations preconditioner resilient to soft and hard faults. in J Dongarra, JM Kunkel & P Balaji (eds), High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9697, Springer Verlag, pp. 469-485, 31st International Conference on High Performance Computing, ISC High Performance 2016, Frankfurt, Germany, 06/19/16. https://doi.org/10.1007/978-3-319-41321-1_24

Scalability of partial differential equations preconditioner resilient to soft and hard faults. / Morris, Karla; Rizzi, Francesco; Sargsyan, Khachik et al.
High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings. ed. / Jack Dongarra; Julian M. Kunkel; Pavan Balaji. Springer Verlag, 2016. p. 469-485 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9697).

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

TY - GEN

T1 - Scalability of partial differential equations preconditioner resilient to soft and hard faults

AU - Morris, Karla

AU - Rizzi, Francesco

AU - Sargsyan, Khachik

AU - Dahlgren, Kathryn

AU - Mycek, Paul

AU - Safta, Cosmin

AU - Le Maître, Olivier

AU - Knio, Omar

AU - Debusschere, Bert

PY - 2016

Y1 - 2016

N2 - We present a resilient domain-decomposition preconditioner for partial differential equations (PDEs). The algorithm reformulates the PDE as a sampling problem, followed by a solution update through data manipulation that is resilient to both soft and hard faults. We discuss an implementation based on a server-client model where all state information is held by the servers, while clients are designed solely as computational units. Servers are assumed to be “sandboxed”, while no assumption is made on the reliability of the clients. We explore the scalability of the algorithm up to ∼12k cores, build an SST/macro skeleton to extrapolate to∼50k cores, and show the resilience under simulated hard and soft faults for a 2D linear Poisson equation.

AB - We present a resilient domain-decomposition preconditioner for partial differential equations (PDEs). The algorithm reformulates the PDE as a sampling problem, followed by a solution update through data manipulation that is resilient to both soft and hard faults. We discuss an implementation based on a server-client model where all state information is held by the servers, while clients are designed solely as computational units. Servers are assumed to be “sandboxed”, while no assumption is made on the reliability of the clients. We explore the scalability of the algorithm up to ∼12k cores, build an SST/macro skeleton to extrapolate to∼50k cores, and show the resilience under simulated hard and soft faults for a 2D linear Poisson equation.

UR - http://www.scopus.com/inward/record.url?scp=84977541671&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-41321-1_24

DO - 10.1007/978-3-319-41321-1_24

M3 - Conference contribution

AN - SCOPUS:84977541671

SN - 9783319413204

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

SP - 469

EP - 485

BT - High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings

A2 - Dongarra, Jack

A2 - Kunkel, Julian M.

A2 - Balaji, Pavan

PB - Springer Verlag

T2 - 31st International Conference on High Performance Computing, ISC High Performance 2016

Y2 - 19 June 2016 through 23 June 2016

ER -

Morris K, Rizzi F, Sargsyan K, Dahlgren K, Mycek P, Safta C et al. Scalability of partial differential equations preconditioner resilient to soft and hard faults. In Dongarra J, Kunkel JM, Balaji P, editors, High Performance Computing - 31st International Conference, ISC High Performance 2016, Proceedings. Springer Verlag. 2016. p. 469-485. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-41321-1_24

Scalability of partial differential equations preconditioner resilient to soft and hard faults

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this