Enabling Resilience in Asynchronous Many-Task Programming Models

Sri Raj Paul; Akihiro Hayashi; Nicole Slattengren; Hemanth Kolla; Matthew Whitlock; Seonmyeong Bak; Keita Teranishi; Jackson Mayo; Vivek Sarkar

doi:10.1007/978-3-030-29400-7_25

Enabling Resilience in Asynchronous Many-Task Programming Models

Sri Raj Paul, Akihiro Hayashi, Nicole Slattengren, Hemanth Kolla, Matthew Whitlock, Seonmyeong Bak, Keita Teranishi, Jackson Mayo, Vivek Sarkar

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

8 Scopus citations

Abstract

Resilience is an imminent issue for next-generation platforms due to projected increases in soft/transient failures as part of the inherent trade-offs among performance, energy, and costs in system design. In this paper, we introduce a comprehensive approach to enabling application-level resilience in Asynchronous Many-Task (AMT) programming models with a focus on remedying Silent Data Corruption (SDC) that can often go undetected by the hardware and OS. Our approach makes it possible for the application programmer to declaratively express resilience attributes with minimal code changes, and to delegate the complexity of efficiently supporting resilience to our runtime system. We have created a prototype implementation of our approach as an extension to the Habanero C/C++ library (HClib), where different resilience techniques including task replay, task replication, algorithm-based fault tolerance (ABFT), and checkpointing are available. Our experimental results show that task replay incurs lower overhead than task replication when an appropriate error checking function is provided. Further, task replay matches the low overhead of ABFT. Our results also demonstrate the ability to combine different resilience schemes. To evaluate the effectiveness of our resilience mechanisms in the presence of errors, we injected synthetic errors at different error rates (1.0%, and 10.0%) and found modest increase in execution times. In summary, the results show that our approach supports efficient and scalable recovery, and that our approach can be used to influence the design of future AMT programming models and runtime systems that aim to integrate first-class support for user-level resilience.

Original language	English
Title of host publication	Euro-Par 2019
Subtitle of host publication	Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings
Editors	Ramin Yahyapour
Publisher	Springer
Pages	346-360
Number of pages	15
ISBN (Print)	9783030293994
DOIs	https://doi.org/10.1007/978-3-030-29400-7_25
State	Published - 2019
Externally published	Yes
Event	25th International European Conference on Parallel and Distributed Computing, Euro-Par 2019 - Göttingen, Germany Duration: Aug 26 2019 → Aug 30 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11725 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	25th International European Conference on Parallel and Distributed Computing, Euro-Par 2019
Country/Territory	Germany
City	Göttingen
Period	08/26/19 → 08/30/19

Funding

Abstract. Resilience is an imminent issue for next-generation platforms due to projected increases in soft/transient failures as part of the inherent trade-offs among performance, energy, and costs in system design. In this paper, we introduce a comprehensive approach to enabling application-level resilience in Asynchronous Many-Task (AMT) programming models with a focus on remedying Silent Data Corruption (SDC) that can often go undetected by the hardware and OS. Our approach makes it possible for the application programmer to declaratively express resilience attributes with minimal code changes, and to delegate the complexity of efficiently supporting resilience to our runtime system. We have created a prototype implementation of our approach as an extension to the Habanero C/C++ library (HClib), where different resilience techniques including task replay, task replication, algorithm-based fault tolerance (ABFT), and checkpointing are available. Our experimental results show that task replay incurs lower overhead than task replication when an appropriate error checking function is provided. Further, task replay matches the low overhead of ABFT. Our results also demonstrate the ability to combine different resilience schemes. To evaluate the effectiveness of our resilience mechanisms in the presence of errors, we injected synthetic errors at different error rates (1.0%, and 10.0%) and found modest increase in execution times. In summary, the results show that our approach supports efficient and scalable recovery, and that our Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration (NNSA) under contract DE-NA0003525. This work was funded by NNSA’s Advanced Simulation and Computing (ASC) Program. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the U.S. Department of Energy or the United States Government.

Keywords

AMT runtimes
Habanero C/C++
Resilience

Access to Document

10.1007/978-3-030-29400-7_25

Cite this

Paul, S. R., Hayashi, A., Slattengren, N., Kolla, H., Whitlock, M., Bak, S., Teranishi, K., Mayo, J., & Sarkar, V. (2019). Enabling Resilience in Asynchronous Many-Task Programming Models. In R. Yahyapour (Ed.), Euro-Par 2019: Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings (pp. 346-360). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11725 LNCS). Springer. https://doi.org/10.1007/978-3-030-29400-7_25

Paul, Sri Raj ; Hayashi, Akihiro ; Slattengren, Nicole et al. / Enabling Resilience in Asynchronous Many-Task Programming Models. Euro-Par 2019: Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings. editor / Ramin Yahyapour. Springer, 2019. pp. 346-360 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b263a01fd70e4d36b895aff086eca610,

title = "Enabling Resilience in Asynchronous Many-Task Programming Models",

abstract = "Resilience is an imminent issue for next-generation platforms due to projected increases in soft/transient failures as part of the inherent trade-offs among performance, energy, and costs in system design. In this paper, we introduce a comprehensive approach to enabling application-level resilience in Asynchronous Many-Task (AMT) programming models with a focus on remedying Silent Data Corruption (SDC) that can often go undetected by the hardware and OS. Our approach makes it possible for the application programmer to declaratively express resilience attributes with minimal code changes, and to delegate the complexity of efficiently supporting resilience to our runtime system. We have created a prototype implementation of our approach as an extension to the Habanero C/C++ library (HClib), where different resilience techniques including task replay, task replication, algorithm-based fault tolerance (ABFT), and checkpointing are available. Our experimental results show that task replay incurs lower overhead than task replication when an appropriate error checking function is provided. Further, task replay matches the low overhead of ABFT. Our results also demonstrate the ability to combine different resilience schemes. To evaluate the effectiveness of our resilience mechanisms in the presence of errors, we injected synthetic errors at different error rates (1.0\%, and 10.0\%) and found modest increase in execution times. In summary, the results show that our approach supports efficient and scalable recovery, and that our approach can be used to influence the design of future AMT programming models and runtime systems that aim to integrate first-class support for user-level resilience.",

keywords = "AMT runtimes, Habanero C/C++, Resilience",

author = "Paul, \{Sri Raj\} and Akihiro Hayashi and Nicole Slattengren and Hemanth Kolla and Matthew Whitlock and Seonmyeong Bak and Keita Teranishi and Jackson Mayo and Vivek Sarkar",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.; 25th International European Conference on Parallel and Distributed Computing, Euro-Par 2019 ; Conference date: 26-08-2019 Through 30-08-2019",

year = "2019",

doi = "10.1007/978-3-030-29400-7\_25",

language = "English",

isbn = "9783030293994",

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

publisher = "Springer",

pages = "346--360",

editor = "Ramin Yahyapour",

booktitle = "Euro-Par 2019",

}

Paul, SR, Hayashi, A, Slattengren, N, Kolla, H, Whitlock, M, Bak, S, Teranishi, K, Mayo, J & Sarkar, V 2019, Enabling Resilience in Asynchronous Many-Task Programming Models. in R Yahyapour (ed.), Euro-Par 2019: Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11725 LNCS, Springer, pp. 346-360, 25th International European Conference on Parallel and Distributed Computing, Euro-Par 2019, Göttingen, Germany, 08/26/19. https://doi.org/10.1007/978-3-030-29400-7_25

Enabling Resilience in Asynchronous Many-Task Programming Models. / Paul, Sri Raj; Hayashi, Akihiro; Slattengren, Nicole et al.
Euro-Par 2019: Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings. ed. / Ramin Yahyapour. Springer, 2019. p. 346-360 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11725 LNCS).

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

TY - GEN

T1 - Enabling Resilience in Asynchronous Many-Task Programming Models

AU - Paul, Sri Raj

AU - Hayashi, Akihiro

AU - Slattengren, Nicole

AU - Kolla, Hemanth

AU - Whitlock, Matthew

AU - Bak, Seonmyeong

AU - Teranishi, Keita

AU - Mayo, Jackson

AU - Sarkar, Vivek

PY - 2019

Y1 - 2019

N2 - Resilience is an imminent issue for next-generation platforms due to projected increases in soft/transient failures as part of the inherent trade-offs among performance, energy, and costs in system design. In this paper, we introduce a comprehensive approach to enabling application-level resilience in Asynchronous Many-Task (AMT) programming models with a focus on remedying Silent Data Corruption (SDC) that can often go undetected by the hardware and OS. Our approach makes it possible for the application programmer to declaratively express resilience attributes with minimal code changes, and to delegate the complexity of efficiently supporting resilience to our runtime system. We have created a prototype implementation of our approach as an extension to the Habanero C/C++ library (HClib), where different resilience techniques including task replay, task replication, algorithm-based fault tolerance (ABFT), and checkpointing are available. Our experimental results show that task replay incurs lower overhead than task replication when an appropriate error checking function is provided. Further, task replay matches the low overhead of ABFT. Our results also demonstrate the ability to combine different resilience schemes. To evaluate the effectiveness of our resilience mechanisms in the presence of errors, we injected synthetic errors at different error rates (1.0%, and 10.0%) and found modest increase in execution times. In summary, the results show that our approach supports efficient and scalable recovery, and that our approach can be used to influence the design of future AMT programming models and runtime systems that aim to integrate first-class support for user-level resilience.

AB - Resilience is an imminent issue for next-generation platforms due to projected increases in soft/transient failures as part of the inherent trade-offs among performance, energy, and costs in system design. In this paper, we introduce a comprehensive approach to enabling application-level resilience in Asynchronous Many-Task (AMT) programming models with a focus on remedying Silent Data Corruption (SDC) that can often go undetected by the hardware and OS. Our approach makes it possible for the application programmer to declaratively express resilience attributes with minimal code changes, and to delegate the complexity of efficiently supporting resilience to our runtime system. We have created a prototype implementation of our approach as an extension to the Habanero C/C++ library (HClib), where different resilience techniques including task replay, task replication, algorithm-based fault tolerance (ABFT), and checkpointing are available. Our experimental results show that task replay incurs lower overhead than task replication when an appropriate error checking function is provided. Further, task replay matches the low overhead of ABFT. Our results also demonstrate the ability to combine different resilience schemes. To evaluate the effectiveness of our resilience mechanisms in the presence of errors, we injected synthetic errors at different error rates (1.0%, and 10.0%) and found modest increase in execution times. In summary, the results show that our approach supports efficient and scalable recovery, and that our approach can be used to influence the design of future AMT programming models and runtime systems that aim to integrate first-class support for user-level resilience.

KW - AMT runtimes

KW - Habanero C/C++

KW - Resilience

UR - https://www.scopus.com/pages/publications/85077127899

U2 - 10.1007/978-3-030-29400-7_25

DO - 10.1007/978-3-030-29400-7_25

M3 - Conference contribution

AN - SCOPUS:85077127899

SN - 9783030293994

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

SP - 346

EP - 360

BT - Euro-Par 2019

A2 - Yahyapour, Ramin

PB - Springer

T2 - 25th International European Conference on Parallel and Distributed Computing, Euro-Par 2019

Y2 - 26 August 2019 through 30 August 2019

ER -

Paul SR, Hayashi A, Slattengren N, Kolla H, Whitlock M, Bak S et al. Enabling Resilience in Asynchronous Many-Task Programming Models. In Yahyapour R, editor, Euro-Par 2019: Parallel Processing - 25th International Conference on Parallel and Distributed Computing, Proceedings. Springer. 2019. p. 346-360. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-29400-7_25

Enabling Resilience in Asynchronous Many-Task Programming Models

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this