AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems

Jie Jia; Yi Liu; Yanke Liu; Yifan Chen; Fang Lin

doi:10.1007/978-3-031-69583-4_24

AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems

Jie Jia^*
, Yi Liu
, Yanke Liu
, Yifan Chen
, Fang Lin

^*Corresponding author for this work

School of Computer Science and Engineering

Beihang University

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

1 Scopus citations

Abstract

With the scaling-up of high-performance computing (HPC) systems, the resilience has become an important challenge. As a widely used resilience technique for HPC systems, checkpointing saves checkpoints of the system during the execution of parallel programs, and in case of failure, recovers the execution of the program from the most recent checkpoint. However, large-scale parallel programs often produce thousands of processes, and result in large-volume simultaneous data-writings on each checkpoint, which impacts the storage as well as the parallel file systems of HPC. To tackle this problem, this paper proposes AdapCK, an I/O-optimization scheme for checkpointing on large-scale HPC systems. AdapCK consists of two main parts: a load-balancing mechanism used for balancing workloads across low-level storage volumes on checkpointing, and a throughput-aware checkpoint-data writing mechanism used for reducing I/O contentions and increasing utilization of I/O-bandwidth. Experiment results show that the AdapCK can reduce the checkpoint time by more than 30%, up to 54.5%.

Original language	English
Title of host publication	Euro-Par 2024
Subtitle of host publication	Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings
Editors	Jesus Carretero, Javier Garcia-Blas, Sameer Shende, Ivona Brandic, Katzalin Olcoz, Martin Schreiber
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	342-355
Number of pages	14
ISBN (Print)	9783031695827
DOIs	https://doi.org/10.1007/978-3-031-69583-4_24
State	Published - 2024
Event	30th International Conference on Parallel and Distributed Computing, Euro-Par 2024 - Madrid, Spain Duration: 26 Aug 2024 → 30 Aug 2024

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14803 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	30th International Conference on Parallel and Distributed Computing, Euro-Par 2024
Country/Territory	Spain
City	Madrid
Period	26/08/24 → 30/08/24

Keywords

Checkpoint
DMTCP
Fault tolerance
High-Performance Computing
Parallel file system

Access to Document

10.1007/978-3-031-69583-4_24

Cite this

Jia, J., Liu, Y., Liu, Y., Chen, Y., & Lin, F. (2024). AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems. In J. Carretero, J. Garcia-Blas, S. Shende, I. Brandic, K. Olcoz, & M. Schreiber (Eds.), Euro-Par 2024: Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings (pp. 342-355). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14803 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-69583-4_24

Jia, Jie ; Liu, Yi ; Liu, Yanke et al. / AdapCK : Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems. Euro-Par 2024: Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings. editor / Jesus Carretero ; Javier Garcia-Blas ; Sameer Shende ; Ivona Brandic ; Katzalin Olcoz ; Martin Schreiber. Springer Science and Business Media Deutschland GmbH, 2024. pp. 342-355 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{5324b1b8e4e74c809e570bedfd83af3a,

title = "AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems",

abstract = "With the scaling-up of high-performance computing (HPC) systems, the resilience has become an important challenge. As a widely used resilience technique for HPC systems, checkpointing saves checkpoints of the system during the execution of parallel programs, and in case of failure, recovers the execution of the program from the most recent checkpoint. However, large-scale parallel programs often produce thousands of processes, and result in large-volume simultaneous data-writings on each checkpoint, which impacts the storage as well as the parallel file systems of HPC. To tackle this problem, this paper proposes AdapCK, an I/O-optimization scheme for checkpointing on large-scale HPC systems. AdapCK consists of two main parts: a load-balancing mechanism used for balancing workloads across low-level storage volumes on checkpointing, and a throughput-aware checkpoint-data writing mechanism used for reducing I/O contentions and increasing utilization of I/O-bandwidth. Experiment results show that the AdapCK can reduce the checkpoint time by more than 30\%, up to 54.5\%.",

keywords = "Checkpoint, DMTCP, Fault tolerance, High-Performance Computing, Parallel file system",

author = "Jie Jia and Yi Liu and Yanke Liu and Yifan Chen and Fang Lin",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.; 30th International Conference on Parallel and Distributed Computing, Euro-Par 2024 ; Conference date: 26-08-2024 Through 30-08-2024",

year = "2024",

doi = "10.1007/978-3-031-69583-4\_24",

language = "英语",

isbn = "9783031695827",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "342--355",

editor = "Jesus Carretero and Javier Garcia-Blas and Sameer Shende and Ivona Brandic and Katzalin Olcoz and Martin Schreiber",

booktitle = "Euro-Par 2024",

address = "德国",

}

Jia, J, Liu, Y, Liu, Y, Chen, Y & Lin, F 2024, AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems. in J Carretero, J Garcia-Blas, S Shende, I Brandic, K Olcoz & M Schreiber (eds), Euro-Par 2024: Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14803 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 342-355, 30th International Conference on Parallel and Distributed Computing, Euro-Par 2024, Madrid, Spain, 26/08/24. https://doi.org/10.1007/978-3-031-69583-4_24

AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems. / Jia, Jie; Liu, Yi; Liu, Yanke et al.
Euro-Par 2024: Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings. ed. / Jesus Carretero; Javier Garcia-Blas; Sameer Shende; Ivona Brandic; Katzalin Olcoz; Martin Schreiber. Springer Science and Business Media Deutschland GmbH, 2024. p. 342-355 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14803 LNCS).

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

TY - GEN

T1 - AdapCK

T2 - 30th International Conference on Parallel and Distributed Computing, Euro-Par 2024

AU - Jia, Jie

AU - Liu, Yi

AU - Liu, Yanke

AU - Chen, Yifan

AU - Lin, Fang

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

PY - 2024

Y1 - 2024

N2 - With the scaling-up of high-performance computing (HPC) systems, the resilience has become an important challenge. As a widely used resilience technique for HPC systems, checkpointing saves checkpoints of the system during the execution of parallel programs, and in case of failure, recovers the execution of the program from the most recent checkpoint. However, large-scale parallel programs often produce thousands of processes, and result in large-volume simultaneous data-writings on each checkpoint, which impacts the storage as well as the parallel file systems of HPC. To tackle this problem, this paper proposes AdapCK, an I/O-optimization scheme for checkpointing on large-scale HPC systems. AdapCK consists of two main parts: a load-balancing mechanism used for balancing workloads across low-level storage volumes on checkpointing, and a throughput-aware checkpoint-data writing mechanism used for reducing I/O contentions and increasing utilization of I/O-bandwidth. Experiment results show that the AdapCK can reduce the checkpoint time by more than 30%, up to 54.5%.

AB - With the scaling-up of high-performance computing (HPC) systems, the resilience has become an important challenge. As a widely used resilience technique for HPC systems, checkpointing saves checkpoints of the system during the execution of parallel programs, and in case of failure, recovers the execution of the program from the most recent checkpoint. However, large-scale parallel programs often produce thousands of processes, and result in large-volume simultaneous data-writings on each checkpoint, which impacts the storage as well as the parallel file systems of HPC. To tackle this problem, this paper proposes AdapCK, an I/O-optimization scheme for checkpointing on large-scale HPC systems. AdapCK consists of two main parts: a load-balancing mechanism used for balancing workloads across low-level storage volumes on checkpointing, and a throughput-aware checkpoint-data writing mechanism used for reducing I/O contentions and increasing utilization of I/O-bandwidth. Experiment results show that the AdapCK can reduce the checkpoint time by more than 30%, up to 54.5%.

KW - Checkpoint

KW - DMTCP

KW - Fault tolerance

KW - High-Performance Computing

KW - Parallel file system

UR - https://www.scopus.com/pages/publications/85202768856

U2 - 10.1007/978-3-031-69583-4_24

DO - 10.1007/978-3-031-69583-4_24

M3 - 会议稿件

AN - SCOPUS:85202768856

SN - 9783031695827

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

SP - 342

EP - 355

BT - Euro-Par 2024

A2 - Carretero, Jesus

A2 - Garcia-Blas, Javier

A2 - Shende, Sameer

A2 - Brandic, Ivona

A2 - Olcoz, Katzalin

A2 - Schreiber, Martin

PB - Springer Science and Business Media Deutschland GmbH

Y2 - 26 August 2024 through 30 August 2024

ER -

Jia J, Liu Y, Liu Y, Chen Y, Lin F. AdapCK: Optimizing I/O for Checkpointing on Large-Scale High Performance Computing Systems. In Carretero J, Garcia-Blas J, Shende S, Brandic I, Olcoz K, Schreiber M, editors, Euro-Par 2024: Parallel Processing - 30th European Conference on Parallel and Distributed Processing, Proceedings. Springer Science and Business Media Deutschland GmbH. 2024. p. 342-355. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-69583-4_24