Dyn-MPI

Supporting MPI on non dedicated clusters

D. Brent Weatherly, David K Lowenthal, Mario Nakazawa, Franklin Lowenthal

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

3 Citations (Scopus)

Abstract

Distributing data is a fundamental problem in implementing efficient distributed-memory parallel programs. The problem becomes more difficult in environments where the participating nodes are not dedicated to a parallel application. We are investigating the data distribution problem in non dedicated environments in the context of explicit message-passing programs. To address this problem, we have designed and implemented an extension to MPI called Dynamic MPI (Dyn-MPI). The key component of Dyn-MPI is its run-time system, which efficiently and automatically redistributes data on the fly when there are changes in the application or the underlying environment. Dyn-MPI supports efficient memory allocation, precise measurement of system load and computation time, and node removal. Performance results show that programs that use Dyn-MPI execute efficiently in non dedicated environments, including up to almost a three-fold improvement compared to programs that do not redistribute data and a 25% improvement over standard adaptive load balancing techniques.

Original languageEnglish (US)
Title of host publicationProceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003
DOIs
StatePublished - 2003
Externally publishedYes
Event2003 ACM/IEEE Conference on Supercomputing, SC 2003 - Phoenix, AZ, United States
Duration: Nov 15 2003Nov 21 2003

Other

Other2003 ACM/IEEE Conference on Supercomputing, SC 2003
CountryUnited States
CityPhoenix, AZ
Period11/15/0311/21/03

Fingerprint

Storage allocation (computer)
Message passing
Resource allocation
Data storage equipment

ASJC Scopus subject areas

  • Software

Cite this

Weatherly, D. B., Lowenthal, D. K., Nakazawa, M., & Lowenthal, F. (2003). Dyn-MPI: Supporting MPI on non dedicated clusters. In Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003 https://doi.org/10.1145/1048935.1050156

Dyn-MPI : Supporting MPI on non dedicated clusters. / Weatherly, D. Brent; Lowenthal, David K; Nakazawa, Mario; Lowenthal, Franklin.

Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003. 2003.

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

Weatherly, DB, Lowenthal, DK, Nakazawa, M & Lowenthal, F 2003, Dyn-MPI: Supporting MPI on non dedicated clusters. in Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003. 2003 ACM/IEEE Conference on Supercomputing, SC 2003, Phoenix, AZ, United States, 11/15/03. https://doi.org/10.1145/1048935.1050156
Weatherly DB, Lowenthal DK, Nakazawa M, Lowenthal F. Dyn-MPI: Supporting MPI on non dedicated clusters. In Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003. 2003 https://doi.org/10.1145/1048935.1050156
Weatherly, D. Brent ; Lowenthal, David K ; Nakazawa, Mario ; Lowenthal, Franklin. / Dyn-MPI : Supporting MPI on non dedicated clusters. Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003. 2003.
@inproceedings{c428fe476e554450925d5998bf044551,
title = "Dyn-MPI: Supporting MPI on non dedicated clusters",
abstract = "Distributing data is a fundamental problem in implementing efficient distributed-memory parallel programs. The problem becomes more difficult in environments where the participating nodes are not dedicated to a parallel application. We are investigating the data distribution problem in non dedicated environments in the context of explicit message-passing programs. To address this problem, we have designed and implemented an extension to MPI called Dynamic MPI (Dyn-MPI). The key component of Dyn-MPI is its run-time system, which efficiently and automatically redistributes data on the fly when there are changes in the application or the underlying environment. Dyn-MPI supports efficient memory allocation, precise measurement of system load and computation time, and node removal. Performance results show that programs that use Dyn-MPI execute efficiently in non dedicated environments, including up to almost a three-fold improvement compared to programs that do not redistribute data and a 25{\%} improvement over standard adaptive load balancing techniques.",
author = "Weatherly, {D. Brent} and Lowenthal, {David K} and Mario Nakazawa and Franklin Lowenthal",
year = "2003",
doi = "10.1145/1048935.1050156",
language = "English (US)",
isbn = "1581136951",
booktitle = "Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003",

}

TY - GEN

T1 - Dyn-MPI

T2 - Supporting MPI on non dedicated clusters

AU - Weatherly, D. Brent

AU - Lowenthal, David K

AU - Nakazawa, Mario

AU - Lowenthal, Franklin

PY - 2003

Y1 - 2003

N2 - Distributing data is a fundamental problem in implementing efficient distributed-memory parallel programs. The problem becomes more difficult in environments where the participating nodes are not dedicated to a parallel application. We are investigating the data distribution problem in non dedicated environments in the context of explicit message-passing programs. To address this problem, we have designed and implemented an extension to MPI called Dynamic MPI (Dyn-MPI). The key component of Dyn-MPI is its run-time system, which efficiently and automatically redistributes data on the fly when there are changes in the application or the underlying environment. Dyn-MPI supports efficient memory allocation, precise measurement of system load and computation time, and node removal. Performance results show that programs that use Dyn-MPI execute efficiently in non dedicated environments, including up to almost a three-fold improvement compared to programs that do not redistribute data and a 25% improvement over standard adaptive load balancing techniques.

AB - Distributing data is a fundamental problem in implementing efficient distributed-memory parallel programs. The problem becomes more difficult in environments where the participating nodes are not dedicated to a parallel application. We are investigating the data distribution problem in non dedicated environments in the context of explicit message-passing programs. To address this problem, we have designed and implemented an extension to MPI called Dynamic MPI (Dyn-MPI). The key component of Dyn-MPI is its run-time system, which efficiently and automatically redistributes data on the fly when there are changes in the application or the underlying environment. Dyn-MPI supports efficient memory allocation, precise measurement of system load and computation time, and node removal. Performance results show that programs that use Dyn-MPI execute efficiently in non dedicated environments, including up to almost a three-fold improvement compared to programs that do not redistribute data and a 25% improvement over standard adaptive load balancing techniques.

UR - http://www.scopus.com/inward/record.url?scp=84877056877&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84877056877&partnerID=8YFLogxK

U2 - 10.1145/1048935.1050156

DO - 10.1145/1048935.1050156

M3 - Conference contribution

SN - 1581136951

SN - 9781581136951

BT - Proceedings of the 2003 ACM/IEEE Conference on Supercomputing, SC 2003

ER -