Plan B: Interruption of ongoing MPI operations to support failure recovery

Aurelien Bouteiller; George Bosilca; Jack J. Dongarra

doi:10.1145/2802658.2802668

Plan B: Interruption of ongoing MPI operations to support failure recovery

Aurelien Bouteiller
, George Bosilca
, Jack J. Dongarra

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

8 Scopus citations

Abstract

Advanced failure recovery strategies in HPC system benefit tremendously from in-place failure recovery, in which the MPI infrastructure can survive process crashes and resume communication services. In this paper we present the rationale behind the specification, and an effective implementation of the Revoke MPI operation. The purpose of the Revoke operation is the propagation of failure knowledge, and the interruption of ongoing, pending communication, under the control of the user. We explain that the Revoke operation can be implemented with a reliable broadcast over the scalable and failure resilient Binomial Graph (BMG) over- lay network. Evaluation at scale, on a Cray XC30 super- computer, demonstrates that the Revoke operation has a small latency, and does not introduce system noise outside of failure recovery periods.

Original language	English
Title of host publication	Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015
Publisher	Association for Computing Machinery
ISBN (Electronic)	9781450337953
DOIs	https://doi.org/10.1145/2802658.2802668
State	Published - Sep 21 2015
Externally published	Yes
Event	22nd European MPI Users' Group Meeting, EuroMPI 2015 - Bordeaux, France Duration: Sep 21 2015 → Sep 23 2015

Publication series

Name	ACM International Conference Proceeding Series
Volume	21-23-September-2015

Conference

Conference	22nd European MPI Users' Group Meeting, EuroMPI 2015
Country/Territory	France
City	Bordeaux
Period	09/21/15 → 09/23/15

Funding

This work is partially supported by the CREST project of the Japan Science and Technology Agency (JST), and by NSF award #1339820.

Access to Document

10.1145/2802658.2802668

Cite this

Bouteiller, A., Bosilca, G., & Dongarra, J. J. (2015). Plan B: Interruption of ongoing MPI operations to support failure recovery. In Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015 Article a11 (ACM International Conference Proceeding Series; Vol. 21-23-September-2015). Association for Computing Machinery. https://doi.org/10.1145/2802658.2802668

@inproceedings{eb01300bcf344083a81cc972cc28b132,

title = "Plan B: Interruption of ongoing MPI operations to support failure recovery",

abstract = "Advanced failure recovery strategies in HPC system benefit tremendously from in-place failure recovery, in which the MPI infrastructure can survive process crashes and resume communication services. In this paper we present the rationale behind the specification, and an effective implementation of the Revoke MPI operation. The purpose of the Revoke operation is the propagation of failure knowledge, and the interruption of ongoing, pending communication, under the control of the user. We explain that the Revoke operation can be implemented with a reliable broadcast over the scalable and failure resilient Binomial Graph (BMG) over- lay network. Evaluation at scale, on a Cray XC30 super- computer, demonstrates that the Revoke operation has a small latency, and does not introduce system noise outside of failure recovery periods.",

author = "Aurelien Bouteiller and George Bosilca and Dongarra, \{Jack J.\}",

note = "Publisher Copyright: {\textcopyright} 2015 ACM.; 22nd European MPI Users' Group Meeting, EuroMPI 2015 ; Conference date: 21-09-2015 Through 23-09-2015",

year = "2015",

month = sep,

day = "21",

doi = "10.1145/2802658.2802668",

language = "English",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

booktitle = "Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015",

}

Bouteiller, A, Bosilca, G & Dongarra, JJ 2015, Plan B: Interruption of ongoing MPI operations to support failure recovery. in Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015., a11, ACM International Conference Proceeding Series, vol. 21-23-September-2015, Association for Computing Machinery, 22nd European MPI Users' Group Meeting, EuroMPI 2015, Bordeaux, France, 09/21/15. https://doi.org/10.1145/2802658.2802668

Plan B: Interruption of ongoing MPI operations to support failure recovery. / Bouteiller, Aurelien; Bosilca, George; Dongarra, Jack J.
Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015. Association for Computing Machinery, 2015. a11 (ACM International Conference Proceeding Series; Vol. 21-23-September-2015).

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

TY - GEN

T1 - Plan B

T2 - 22nd European MPI Users' Group Meeting, EuroMPI 2015

AU - Bouteiller, Aurelien

AU - Bosilca, George

AU - Dongarra, Jack J.

PY - 2015/9/21

Y1 - 2015/9/21

N2 - Advanced failure recovery strategies in HPC system benefit tremendously from in-place failure recovery, in which the MPI infrastructure can survive process crashes and resume communication services. In this paper we present the rationale behind the specification, and an effective implementation of the Revoke MPI operation. The purpose of the Revoke operation is the propagation of failure knowledge, and the interruption of ongoing, pending communication, under the control of the user. We explain that the Revoke operation can be implemented with a reliable broadcast over the scalable and failure resilient Binomial Graph (BMG) over- lay network. Evaluation at scale, on a Cray XC30 super- computer, demonstrates that the Revoke operation has a small latency, and does not introduce system noise outside of failure recovery periods.

AB - Advanced failure recovery strategies in HPC system benefit tremendously from in-place failure recovery, in which the MPI infrastructure can survive process crashes and resume communication services. In this paper we present the rationale behind the specification, and an effective implementation of the Revoke MPI operation. The purpose of the Revoke operation is the propagation of failure knowledge, and the interruption of ongoing, pending communication, under the control of the user. We explain that the Revoke operation can be implemented with a reliable broadcast over the scalable and failure resilient Binomial Graph (BMG) over- lay network. Evaluation at scale, on a Cray XC30 super- computer, demonstrates that the Revoke operation has a small latency, and does not introduce system noise outside of failure recovery periods.

UR - https://www.scopus.com/pages/publications/84983447572

U2 - 10.1145/2802658.2802668

DO - 10.1145/2802658.2802668

M3 - Conference contribution

AN - SCOPUS:84983447572

T3 - ACM International Conference Proceeding Series

BT - Proceedings of the 22nd European MPI Users' Group Meeting, EuroMPI 2015

PB - Association for Computing Machinery

Y2 - 21 September 2015 through 23 September 2015

ER -

Plan B: Interruption of ongoing MPI operations to support failure recovery

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

Cite this