The design of advanced communication to reduce memory usage for exa-scale systems

Shinji Sumimoto, Yuichiro Ajima, Kazushige Saga, Takafumi Nose, Naoyuki Shida, Takeshi Nanri

研究成果: 著書/レポートタイプへの貢献会議での発言

2 引用 (Scopus)

抄録

Current MPI (Message Passing Interface) communication libraries require larger memories in proportion of the number of processes, and can not be used for exa-scale systems. This paper proposes a global memory based communication design to reduce memory usage for exa-scale communication. To realize exa-scale communication, we propose true global memory based communication primitives called Advanced Communication Primitives (ACPs). ACPs provide global address, which is able to use remote atomic memory operations on the global memory, RDMA (Remote Direct Memory Access) based remote memory copy operation, global heap allocator and global data libraries. ACPs are different from the other communication libraries because ACPs are global memory based so that house keeping memories can be distributed to other processes and programmers explicitly consider memory usage by using ACPs. The preliminary result of memory usage by ACPs is 70 MB on one million processes.

元の言語英語
ホスト出版物のタイトルHigh Performance Computing for Computational Science
ホスト出版物のサブタイトルVECPAR 2016 - 12th International Conference, Revised Selected Papers
編集者Ines Dutra, Rui Camacho, Jorge Barbosa, Osni Marques
出版者Springer Verlag
ページ149-161
ページ数13
ISBN(印刷物)9783319619811
DOI
出版物ステータス出版済み - 1 1 2017
イベント12th International Conference on High Performance Computing for Computational Science, VECPAR 2016 - Porto, ポルトガル
継続期間: 6 28 20166 30 2016

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
10150 LNCS
ISSN(印刷物)0302-9743
ISSN(電子版)1611-3349

その他

その他12th International Conference on High Performance Computing for Computational Science, VECPAR 2016
ポルトガル
Porto
期間6/28/166/30/16

Fingerprint

Computer systems
Data storage equipment
Communication
Design
Message Passing Interface
Heap
Message passing
Interfaces (computer)
Proportion

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

これを引用

Sumimoto, S., Ajima, Y., Saga, K., Nose, T., Shida, N., & Nanri, T. (2017). The design of advanced communication to reduce memory usage for exa-scale systems. : I. Dutra, R. Camacho, J. Barbosa, & O. Marques (版), High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers (pp. 149-161). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 巻数 10150 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-61982-8_15

The design of advanced communication to reduce memory usage for exa-scale systems. / Sumimoto, Shinji; Ajima, Yuichiro; Saga, Kazushige; Nose, Takafumi; Shida, Naoyuki; Nanri, Takeshi.

High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers. 版 / Ines Dutra; Rui Camacho; Jorge Barbosa; Osni Marques. Springer Verlag, 2017. p. 149-161 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 巻 10150 LNCS).

研究成果: 著書/レポートタイプへの貢献会議での発言

Sumimoto, S, Ajima, Y, Saga, K, Nose, T, Shida, N & Nanri, T 2017, The design of advanced communication to reduce memory usage for exa-scale systems. : I Dutra, R Camacho, J Barbosa & O Marques (版), High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 巻. 10150 LNCS, Springer Verlag, pp. 149-161, 12th International Conference on High Performance Computing for Computational Science, VECPAR 2016, Porto, ポルトガル, 6/28/16. https://doi.org/10.1007/978-3-319-61982-8_15
Sumimoto S, Ajima Y, Saga K, Nose T, Shida N, Nanri T. The design of advanced communication to reduce memory usage for exa-scale systems. : Dutra I, Camacho R, Barbosa J, Marques O, 編集者, High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers. Springer Verlag. 2017. p. 149-161. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-61982-8_15
Sumimoto, Shinji ; Ajima, Yuichiro ; Saga, Kazushige ; Nose, Takafumi ; Shida, Naoyuki ; Nanri, Takeshi. / The design of advanced communication to reduce memory usage for exa-scale systems. High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers. 編集者 / Ines Dutra ; Rui Camacho ; Jorge Barbosa ; Osni Marques. Springer Verlag, 2017. pp. 149-161 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{6be2f05aa4aa4fd99fa7cf34d74e0cb4,
title = "The design of advanced communication to reduce memory usage for exa-scale systems",
abstract = "Current MPI (Message Passing Interface) communication libraries require larger memories in proportion of the number of processes, and can not be used for exa-scale systems. This paper proposes a global memory based communication design to reduce memory usage for exa-scale communication. To realize exa-scale communication, we propose true global memory based communication primitives called Advanced Communication Primitives (ACPs). ACPs provide global address, which is able to use remote atomic memory operations on the global memory, RDMA (Remote Direct Memory Access) based remote memory copy operation, global heap allocator and global data libraries. ACPs are different from the other communication libraries because ACPs are global memory based so that house keeping memories can be distributed to other processes and programmers explicitly consider memory usage by using ACPs. The preliminary result of memory usage by ACPs is 70 MB on one million processes.",
author = "Shinji Sumimoto and Yuichiro Ajima and Kazushige Saga and Takafumi Nose and Naoyuki Shida and Takeshi Nanri",
year = "2017",
month = "1",
day = "1",
doi = "10.1007/978-3-319-61982-8_15",
language = "English",
isbn = "9783319619811",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "149--161",
editor = "Ines Dutra and Rui Camacho and Jorge Barbosa and Osni Marques",
booktitle = "High Performance Computing for Computational Science",
address = "Germany",

}

TY - GEN

T1 - The design of advanced communication to reduce memory usage for exa-scale systems

AU - Sumimoto, Shinji

AU - Ajima, Yuichiro

AU - Saga, Kazushige

AU - Nose, Takafumi

AU - Shida, Naoyuki

AU - Nanri, Takeshi

PY - 2017/1/1

Y1 - 2017/1/1

N2 - Current MPI (Message Passing Interface) communication libraries require larger memories in proportion of the number of processes, and can not be used for exa-scale systems. This paper proposes a global memory based communication design to reduce memory usage for exa-scale communication. To realize exa-scale communication, we propose true global memory based communication primitives called Advanced Communication Primitives (ACPs). ACPs provide global address, which is able to use remote atomic memory operations on the global memory, RDMA (Remote Direct Memory Access) based remote memory copy operation, global heap allocator and global data libraries. ACPs are different from the other communication libraries because ACPs are global memory based so that house keeping memories can be distributed to other processes and programmers explicitly consider memory usage by using ACPs. The preliminary result of memory usage by ACPs is 70 MB on one million processes.

AB - Current MPI (Message Passing Interface) communication libraries require larger memories in proportion of the number of processes, and can not be used for exa-scale systems. This paper proposes a global memory based communication design to reduce memory usage for exa-scale communication. To realize exa-scale communication, we propose true global memory based communication primitives called Advanced Communication Primitives (ACPs). ACPs provide global address, which is able to use remote atomic memory operations on the global memory, RDMA (Remote Direct Memory Access) based remote memory copy operation, global heap allocator and global data libraries. ACPs are different from the other communication libraries because ACPs are global memory based so that house keeping memories can be distributed to other processes and programmers explicitly consider memory usage by using ACPs. The preliminary result of memory usage by ACPs is 70 MB on one million processes.

UR - http://www.scopus.com/inward/record.url?scp=85026726518&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85026726518&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-61982-8_15

DO - 10.1007/978-3-319-61982-8_15

M3 - Conference contribution

AN - SCOPUS:85026726518

SN - 9783319619811

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

SP - 149

EP - 161

BT - High Performance Computing for Computational Science

A2 - Dutra, Ines

A2 - Camacho, Rui

A2 - Barbosa, Jorge

A2 - Marques, Osni

PB - Springer Verlag

ER -