DS-CUDA

A middleware to use many GPUs in the cloud environment

Minoru Oikawa, Atsushi Kawai, Kentaro Nomura, Kenji Yasuoka, Kazuyuki Yoshikawa, Tetsu Narumi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

38 Citations (Scopus)

Abstract

GPGPU (General-purpose computing on graphics processing units) has several difficulties when used in cloud environment, such as narrow bandwidth, higher cost, and lower security, compared with computation using only CPUs. Most high performance computing applications require huge communication between nodes, and do not fit a cloud environment, since network topology and its bandwidth are not fixed and they affect the performance of the application program. However, there are some applications for which little communication is needed, such as molecular dynamics (MD) simulation with the replica exchange method (REM). For such applications, we propose DS-CUDA (Distributed-shared compute unified device architecture), a middleware to use many GPUs in a cloud environment with lower cost and higher security. It virtualizes GPUs in a cloud such that they appear to be locally installed GPUs in a client machine. Its redundant mechanism ensures reliable calculation with consumer GPUs, which reduce the cost greatly. It also enhances the security level since no data except command and data for GPUs are stored in the cloud side. REM-MD simulation with 64 GPUs showed 58 and 36 times more speed than a locally-installed GPU via InfiniBand and the Internet, respectively.

Original languageEnglish
Title of host publicationProceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012
Pages1207-1214
Number of pages8
DOIs
Publication statusPublished - 2012
Event2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012 - Salt Lake City, UT, United States
Duration: 2012 Nov 102012 Nov 16

Other

Other2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012
CountryUnited States
CitySalt Lake City, UT
Period12/11/1012/11/16

Fingerprint

Middleware
Molecular dynamics
Bandwidth
Costs
Graphics processing unit
Communication
Computer simulation
Application programs
Program processors
Topology
Internet

Keywords

  • Clouds
  • Clustering methods
  • High performance computing
  • Molecular computing

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Science Applications
  • Software

Cite this

Oikawa, M., Kawai, A., Nomura, K., Yasuoka, K., Yoshikawa, K., & Narumi, T. (2012). DS-CUDA: A middleware to use many GPUs in the cloud environment. In Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012 (pp. 1207-1214). [6495928] https://doi.org/10.1109/SC.Companion.2012.146

DS-CUDA : A middleware to use many GPUs in the cloud environment. / Oikawa, Minoru; Kawai, Atsushi; Nomura, Kentaro; Yasuoka, Kenji; Yoshikawa, Kazuyuki; Narumi, Tetsu.

Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012. 2012. p. 1207-1214 6495928.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Oikawa, M, Kawai, A, Nomura, K, Yasuoka, K, Yoshikawa, K & Narumi, T 2012, DS-CUDA: A middleware to use many GPUs in the cloud environment. in Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012., 6495928, pp. 1207-1214, 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012, Salt Lake City, UT, United States, 12/11/10. https://doi.org/10.1109/SC.Companion.2012.146
Oikawa M, Kawai A, Nomura K, Yasuoka K, Yoshikawa K, Narumi T. DS-CUDA: A middleware to use many GPUs in the cloud environment. In Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012. 2012. p. 1207-1214. 6495928 https://doi.org/10.1109/SC.Companion.2012.146
Oikawa, Minoru ; Kawai, Atsushi ; Nomura, Kentaro ; Yasuoka, Kenji ; Yoshikawa, Kazuyuki ; Narumi, Tetsu. / DS-CUDA : A middleware to use many GPUs in the cloud environment. Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012. 2012. pp. 1207-1214
@inproceedings{04b526fac09a4763990f402e72b90452,
title = "DS-CUDA: A middleware to use many GPUs in the cloud environment",
abstract = "GPGPU (General-purpose computing on graphics processing units) has several difficulties when used in cloud environment, such as narrow bandwidth, higher cost, and lower security, compared with computation using only CPUs. Most high performance computing applications require huge communication between nodes, and do not fit a cloud environment, since network topology and its bandwidth are not fixed and they affect the performance of the application program. However, there are some applications for which little communication is needed, such as molecular dynamics (MD) simulation with the replica exchange method (REM). For such applications, we propose DS-CUDA (Distributed-shared compute unified device architecture), a middleware to use many GPUs in a cloud environment with lower cost and higher security. It virtualizes GPUs in a cloud such that they appear to be locally installed GPUs in a client machine. Its redundant mechanism ensures reliable calculation with consumer GPUs, which reduce the cost greatly. It also enhances the security level since no data except command and data for GPUs are stored in the cloud side. REM-MD simulation with 64 GPUs showed 58 and 36 times more speed than a locally-installed GPU via InfiniBand and the Internet, respectively.",
keywords = "Clouds, Clustering methods, High performance computing, Molecular computing",
author = "Minoru Oikawa and Atsushi Kawai and Kentaro Nomura and Kenji Yasuoka and Kazuyuki Yoshikawa and Tetsu Narumi",
year = "2012",
doi = "10.1109/SC.Companion.2012.146",
language = "English",
isbn = "9780769549569",
pages = "1207--1214",
booktitle = "Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012",

}

TY - GEN

T1 - DS-CUDA

T2 - A middleware to use many GPUs in the cloud environment

AU - Oikawa, Minoru

AU - Kawai, Atsushi

AU - Nomura, Kentaro

AU - Yasuoka, Kenji

AU - Yoshikawa, Kazuyuki

AU - Narumi, Tetsu

PY - 2012

Y1 - 2012

N2 - GPGPU (General-purpose computing on graphics processing units) has several difficulties when used in cloud environment, such as narrow bandwidth, higher cost, and lower security, compared with computation using only CPUs. Most high performance computing applications require huge communication between nodes, and do not fit a cloud environment, since network topology and its bandwidth are not fixed and they affect the performance of the application program. However, there are some applications for which little communication is needed, such as molecular dynamics (MD) simulation with the replica exchange method (REM). For such applications, we propose DS-CUDA (Distributed-shared compute unified device architecture), a middleware to use many GPUs in a cloud environment with lower cost and higher security. It virtualizes GPUs in a cloud such that they appear to be locally installed GPUs in a client machine. Its redundant mechanism ensures reliable calculation with consumer GPUs, which reduce the cost greatly. It also enhances the security level since no data except command and data for GPUs are stored in the cloud side. REM-MD simulation with 64 GPUs showed 58 and 36 times more speed than a locally-installed GPU via InfiniBand and the Internet, respectively.

AB - GPGPU (General-purpose computing on graphics processing units) has several difficulties when used in cloud environment, such as narrow bandwidth, higher cost, and lower security, compared with computation using only CPUs. Most high performance computing applications require huge communication between nodes, and do not fit a cloud environment, since network topology and its bandwidth are not fixed and they affect the performance of the application program. However, there are some applications for which little communication is needed, such as molecular dynamics (MD) simulation with the replica exchange method (REM). For such applications, we propose DS-CUDA (Distributed-shared compute unified device architecture), a middleware to use many GPUs in a cloud environment with lower cost and higher security. It virtualizes GPUs in a cloud such that they appear to be locally installed GPUs in a client machine. Its redundant mechanism ensures reliable calculation with consumer GPUs, which reduce the cost greatly. It also enhances the security level since no data except command and data for GPUs are stored in the cloud side. REM-MD simulation with 64 GPUs showed 58 and 36 times more speed than a locally-installed GPU via InfiniBand and the Internet, respectively.

KW - Clouds

KW - Clustering methods

KW - High performance computing

KW - Molecular computing

UR - http://www.scopus.com/inward/record.url?scp=84876533447&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84876533447&partnerID=8YFLogxK

U2 - 10.1109/SC.Companion.2012.146

DO - 10.1109/SC.Companion.2012.146

M3 - Conference contribution

SN - 9780769549569

SP - 1207

EP - 1214

BT - Proceedings - 2012 SC Companion: High Performance Computing, Networking Storage and Analysis, SCC 2012

ER -