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

doi:10.1007/978-3-319-61982-8_15

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

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

2 Citations (Scopus)

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.

Original language	English
Title of host publication	High Performance Computing for Computational Science
Subtitle of host publication	VECPAR 2016 - 12th International Conference, Revised Selected Papers
Editors	Ines Dutra, Rui Camacho, Jorge Barbosa, Osni Marques
Publisher	Springer Verlag
Pages	149-161
Number of pages	13
ISBN (Print)	9783319619811
DOIs	https://doi.org/10.1007/978-3-319-61982-8_15
Publication status	Published - 2017
Event	12th International Conference on High Performance Computing for Computational Science, VECPAR 2016 - Porto, Portugal Duration: Jun 28 2016 → Jun 30 2016

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10150 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Other

Other	12th International Conference on High Performance Computing for Computational Science, VECPAR 2016
Country/Territory	Portugal
City	Porto
Period	6/28/16 → 6/30/16

All Science Journal Classification (ASJC) codes

Theoretical Computer Science
Computer Science(all)

Access to Document

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

Cite this

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. In I. Dutra, R. Camacho, J. Barbosa, & O. Marques (Eds.), 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); Vol. 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 et al.
High Performance Computing for Computational Science: VECPAR 2016 - 12th International Conference, Revised Selected Papers. ed. / 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); Vol. 10150 LNCS).

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

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. in I Dutra, R Camacho, J Barbosa & O Marques (eds), 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), vol. 10150 LNCS, Springer Verlag, pp. 149-161, 12th International Conference on High Performance Computing for Computational Science, VECPAR 2016, Porto, Portugal, 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. In Dutra I, Camacho R, Barbosa J, Marques O, editors, 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)). doi: 10.1007/978-3-319-61982-8_15

Sumimoto, Shinji ; Ajima, Yuichiro ; Saga, Kazushige et al. / 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. editor / 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",

note = "Funding Information: This research was supported by JST, CREST. Publisher Copyright: {\textcopyright} Springer International Publishing AG 2017.; 12th International Conference on High Performance Computing for Computational Science, VECPAR 2016 ; Conference date: 28-06-2016 Through 30-06-2016",

year = "2017",

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

N1 - Funding Information: This research was supported by JST, CREST. Publisher Copyright: © Springer International Publishing AG 2017.

PY - 2017

Y1 - 2017

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

T2 - 12th International Conference on High Performance Computing for Computational Science, VECPAR 2016

Y2 - 28 June 2016 through 30 June 2016

ER -

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

Abstract

Publication series

Other

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this