Efficient Classification of Supercomputer Failures Using Neuromorphic Computing

Prasanna Date; Christopher D. Carothers; James A. Hendler; Malik Magdon-Ismail

doi:10.1109/SSCI.2018.8628946

Efficient Classification of Supercomputer Failures Using Neuromorphic Computing

Prasanna Date
, Christopher D. Carothers
, James A. Hendler
, Malik Magdon-Ismail

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

16 Scopus citations

Abstract

Today's petascale supercomputers are comprised of ten's of thousands of compute nodes. Failures on these massive machines are a growing problem as the time for a single compute node to fail is shrinking. Ideally, the job scheduler would like the capability to predict node failures ahead of time in order to minimize the impact of node failures on overall job throughput. However, due to the tight power constraints of future systems, the online modeling of real-time error data must be accomplished using as little power as possible. To this end, the IBM TrueNorth Neurosynaptic System is used to create a Spiking Neural Network (SNN) model of supercomputer failure data and the classification accuracy of this model is compared to other Machine Learning (ML) and Deep Learning (DL) techniques. It is observed that the TrueNorth failure classification model yields a training accuracy of 99.41%, validation accuracy of 98.12% and testing accuracy of 99.80% and outperforms other machine learning and deep learning approaches. Moreover, the TrueNorth SNN consumes five orders of magnitude less power than the other ML/DL approaches during the testing phase. Additionally, it is observed that all ML/DL approaches investigated as part of this study are able to produce accurate models of the supercomputer system failure data.

Original language	English
Title of host publication	Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018
Editors	Suresh Sundaram
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	242-249
Number of pages	8
ISBN (Electronic)	9781538692769
DOIs	https://doi.org/10.1109/SSCI.2018.8628946
State	Published - Jul 2 2018
Externally published	Yes
Event	8th IEEE Symposium Series on Computational Intelligence, SSCI 2018 - Bangalore, India Duration: Nov 18 2018 → Nov 21 2018

Publication series

Name	Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018

Conference

Conference	8th IEEE Symposium Series on Computational Intelligence, SSCI 2018
Country/Territory	India
City	Bangalore
Period	11/18/18 → 11/21/18

Funding

VI. ACKNOWLEDGMENT This work was funded by the Air Force Research Lab (AFRL), USA (Award Number: FA8750-15-2-0078). The authors are grateful to IBM Research for lending the TrueNorth chip, development kit and organizing the TrueNorth Boot-camp, and would specifically like to thank Dr. Ben Shaw for getting our paper reviewed by IBM. Dr. Catherine Schumann of the Oak Ridge National Laboratory was very generous to provide her feedback on the paper. The authors would further like to thank Rensselaer Polytechnic Institute (RPI) and the Center for Computational Innovation (CCI) at RPI.

Keywords

Deep Learning
Machine Learning
Neuromorphic Computing
Supercomputer Failures

Access to Document

10.1109/SSCI.2018.8628946

Cite this

Date, P., Carothers, C. D., Hendler, J. A., & Magdon-Ismail, M. (2018). Efficient Classification of Supercomputer Failures Using Neuromorphic Computing. In S. Sundaram (Ed.), Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018 (pp. 242-249). Article 8628946 (Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SSCI.2018.8628946

Date, Prasanna ; Carothers, Christopher D. ; Hendler, James A. et al. / Efficient Classification of Supercomputer Failures Using Neuromorphic Computing. Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. editor / Suresh Sundaram. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 242-249 (Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018).

@inproceedings{c3eb86feb0184a93a3d6103ac982c047,

title = "Efficient Classification of Supercomputer Failures Using Neuromorphic Computing",

abstract = "Today's petascale supercomputers are comprised of ten's of thousands of compute nodes. Failures on these massive machines are a growing problem as the time for a single compute node to fail is shrinking. Ideally, the job scheduler would like the capability to predict node failures ahead of time in order to minimize the impact of node failures on overall job throughput. However, due to the tight power constraints of future systems, the online modeling of real-time error data must be accomplished using as little power as possible. To this end, the IBM TrueNorth Neurosynaptic System is used to create a Spiking Neural Network (SNN) model of supercomputer failure data and the classification accuracy of this model is compared to other Machine Learning (ML) and Deep Learning (DL) techniques. It is observed that the TrueNorth failure classification model yields a training accuracy of 99.41\%, validation accuracy of 98.12\% and testing accuracy of 99.80\% and outperforms other machine learning and deep learning approaches. Moreover, the TrueNorth SNN consumes five orders of magnitude less power than the other ML/DL approaches during the testing phase. Additionally, it is observed that all ML/DL approaches investigated as part of this study are able to produce accurate models of the supercomputer system failure data.",

keywords = "Deep Learning, Machine Learning, Neuromorphic Computing, Supercomputer Failures",

author = "Prasanna Date and Carothers, \{Christopher D.\} and Hendler, \{James A.\} and Malik Magdon-Ismail",

note = "Publisher Copyright: {\textcopyright} 2018 IEEE.; 8th IEEE Symposium Series on Computational Intelligence, SSCI 2018 ; Conference date: 18-11-2018 Through 21-11-2018",

year = "2018",

month = jul,

day = "2",

doi = "10.1109/SSCI.2018.8628946",

language = "English",

series = "Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "242--249",

editor = "Suresh Sundaram",

booktitle = "Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018",

}

Date, P, Carothers, CD, Hendler, JA & Magdon-Ismail, M 2018, Efficient Classification of Supercomputer Failures Using Neuromorphic Computing. in S Sundaram (ed.), Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018., 8628946, Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018, Institute of Electrical and Electronics Engineers Inc., pp. 242-249, 8th IEEE Symposium Series on Computational Intelligence, SSCI 2018, Bangalore, India, 11/18/18. https://doi.org/10.1109/SSCI.2018.8628946

Efficient Classification of Supercomputer Failures Using Neuromorphic Computing. / Date, Prasanna; Carothers, Christopher D.; Hendler, James A. et al.
Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. ed. / Suresh Sundaram. Institute of Electrical and Electronics Engineers Inc., 2018. p. 242-249 8628946 (Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018).

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

TY - GEN

T1 - Efficient Classification of Supercomputer Failures Using Neuromorphic Computing

AU - Date, Prasanna

AU - Carothers, Christopher D.

AU - Hendler, James A.

AU - Magdon-Ismail, Malik

PY - 2018/7/2

Y1 - 2018/7/2

N2 - Today's petascale supercomputers are comprised of ten's of thousands of compute nodes. Failures on these massive machines are a growing problem as the time for a single compute node to fail is shrinking. Ideally, the job scheduler would like the capability to predict node failures ahead of time in order to minimize the impact of node failures on overall job throughput. However, due to the tight power constraints of future systems, the online modeling of real-time error data must be accomplished using as little power as possible. To this end, the IBM TrueNorth Neurosynaptic System is used to create a Spiking Neural Network (SNN) model of supercomputer failure data and the classification accuracy of this model is compared to other Machine Learning (ML) and Deep Learning (DL) techniques. It is observed that the TrueNorth failure classification model yields a training accuracy of 99.41%, validation accuracy of 98.12% and testing accuracy of 99.80% and outperforms other machine learning and deep learning approaches. Moreover, the TrueNorth SNN consumes five orders of magnitude less power than the other ML/DL approaches during the testing phase. Additionally, it is observed that all ML/DL approaches investigated as part of this study are able to produce accurate models of the supercomputer system failure data.

AB - Today's petascale supercomputers are comprised of ten's of thousands of compute nodes. Failures on these massive machines are a growing problem as the time for a single compute node to fail is shrinking. Ideally, the job scheduler would like the capability to predict node failures ahead of time in order to minimize the impact of node failures on overall job throughput. However, due to the tight power constraints of future systems, the online modeling of real-time error data must be accomplished using as little power as possible. To this end, the IBM TrueNorth Neurosynaptic System is used to create a Spiking Neural Network (SNN) model of supercomputer failure data and the classification accuracy of this model is compared to other Machine Learning (ML) and Deep Learning (DL) techniques. It is observed that the TrueNorth failure classification model yields a training accuracy of 99.41%, validation accuracy of 98.12% and testing accuracy of 99.80% and outperforms other machine learning and deep learning approaches. Moreover, the TrueNorth SNN consumes five orders of magnitude less power than the other ML/DL approaches during the testing phase. Additionally, it is observed that all ML/DL approaches investigated as part of this study are able to produce accurate models of the supercomputer system failure data.

KW - Deep Learning

KW - Machine Learning

KW - Neuromorphic Computing

KW - Supercomputer Failures

UR - https://www.scopus.com/pages/publications/85062782411

U2 - 10.1109/SSCI.2018.8628946

DO - 10.1109/SSCI.2018.8628946

M3 - Conference contribution

AN - SCOPUS:85062782411

T3 - Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018

SP - 242

EP - 249

BT - Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018

A2 - Sundaram, Suresh

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 8th IEEE Symposium Series on Computational Intelligence, SSCI 2018

Y2 - 18 November 2018 through 21 November 2018

ER -

Date P, Carothers CD, Hendler JA, Magdon-Ismail M. Efficient Classification of Supercomputer Failures Using Neuromorphic Computing. In Sundaram S, editor, Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 242-249. 8628946. (Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018). doi: 10.1109/SSCI.2018.8628946

Efficient Classification of Supercomputer Failures Using Neuromorphic Computing

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this