Preliminary Study on Message Aggregation Optimizations for Energy Savings in PGAS Models

Oscar Hernandez; Aaron Welch; Wendy Poole; Stephen Poole

doi:10.1007/978-3-031-97492-2_12

Preliminary Study on Message Aggregation Optimizations for Energy Savings in PGAS Models

Oscar Hernandez, Aaron Welch, Wendy Poole, Stephen Poole

Intelligent Systems and Facilities

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

Abstract

As Moore’s law has slowed down, we need to explore opportunities to save energy and power in parallel applications, especially during data movement. This is particularly true within the partitioned global address space (PGAS) model, where there is great potential for energy savings across distributed data accesses on large exascale computing systems. This paper explores the potential of message aggregation strategies within PGAS models, specifically focusing on OpenSHMEM, to improve energy efficiency on the CPU and memory of a node. Using the conveyor library, which aggregates small messages for network-efficient communication, we compare its performance gains in execution times against the energy reductions achieved. We compare applications from the bale effort as implemented through either atomic, get, and put or conveyor approaches on the Frontier supercomputer. Our preliminary results show significant improvements in both performance and energy consumption. These findings suggest that message aggregation can play an important role in addressing the challenges of PGAS energy consumption in modern HPC systems.

Original language	English
Title of host publication	Lecture Notes in Computer Science
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	114-120
Number of pages	7
DOIs	https://doi.org/10.1007/978-3-031-97492-2_12
State	Published - 2025

Publication series

Name	Lecture Notes in Computer Science
Volume	LNCS 14564
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Funding

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan). This research used the Frontier and Andes resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC05-00OR22725. This work was funded through Strategic Partnership Projects Funding Office via Los Alamos National Laboratory with IAN 619215901 on the project “OpenSHMEM-Standardized API for parallel programming in the Partitioned Global Address Space”.

Keywords

energy efficiency
message aggregation
OpenSHMEM
PGAS programming models

Access to Document

10.1007/978-3-031-97492-2_12

Cite this

@inbook{c6aaf1fb88d14146a1e1b9a1c982ffe9,

title = "Preliminary Study on Message Aggregation Optimizations for Energy Savings in PGAS Models",

abstract = "As Moore{\textquoteright}s law has slowed down, we need to explore opportunities to save energy and power in parallel applications, especially during data movement. This is particularly true within the partitioned global address space (PGAS) model, where there is great potential for energy savings across distributed data accesses on large exascale computing systems. This paper explores the potential of message aggregation strategies within PGAS models, specifically focusing on OpenSHMEM, to improve energy efficiency on the CPU and memory of a node. Using the conveyor library, which aggregates small messages for network-efficient communication, we compare its performance gains in execution times against the energy reductions achieved. We compare applications from the bale effort as implemented through either atomic, get, and put or conveyor approaches on the Frontier supercomputer. Our preliminary results show significant improvements in both performance and energy consumption. These findings suggest that message aggregation can play an important role in addressing the challenges of PGAS energy consumption in modern HPC systems.",

keywords = "energy efficiency, message aggregation, OpenSHMEM, PGAS programming models",

author = "Oscar Hernandez and Aaron Welch and Wendy Poole and Stephen Poole",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.",

year = "2025",

doi = "10.1007/978-3-031-97492-2\_12",

language = "English",

series = "Lecture Notes in Computer Science",

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

pages = "114--120",

booktitle = "Lecture Notes in Computer Science",

}

Preliminary Study on Message Aggregation Optimizations for Energy Savings in PGAS Models. / Hernandez, Oscar; Welch, Aaron; Poole, Wendy et al.
Lecture Notes in Computer Science. Springer Science and Business Media Deutschland GmbH, 2025. p. 114-120 (Lecture Notes in Computer Science; Vol. LNCS 14564).

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

TY - CHAP

T1 - Preliminary Study on Message Aggregation Optimizations for Energy Savings in PGAS Models

AU - Hernandez, Oscar

AU - Welch, Aaron

AU - Poole, Wendy

AU - Poole, Stephen

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

PY - 2025

Y1 - 2025

N2 - As Moore’s law has slowed down, we need to explore opportunities to save energy and power in parallel applications, especially during data movement. This is particularly true within the partitioned global address space (PGAS) model, where there is great potential for energy savings across distributed data accesses on large exascale computing systems. This paper explores the potential of message aggregation strategies within PGAS models, specifically focusing on OpenSHMEM, to improve energy efficiency on the CPU and memory of a node. Using the conveyor library, which aggregates small messages for network-efficient communication, we compare its performance gains in execution times against the energy reductions achieved. We compare applications from the bale effort as implemented through either atomic, get, and put or conveyor approaches on the Frontier supercomputer. Our preliminary results show significant improvements in both performance and energy consumption. These findings suggest that message aggregation can play an important role in addressing the challenges of PGAS energy consumption in modern HPC systems.

AB - As Moore’s law has slowed down, we need to explore opportunities to save energy and power in parallel applications, especially during data movement. This is particularly true within the partitioned global address space (PGAS) model, where there is great potential for energy savings across distributed data accesses on large exascale computing systems. This paper explores the potential of message aggregation strategies within PGAS models, specifically focusing on OpenSHMEM, to improve energy efficiency on the CPU and memory of a node. Using the conveyor library, which aggregates small messages for network-efficient communication, we compare its performance gains in execution times against the energy reductions achieved. We compare applications from the bale effort as implemented through either atomic, get, and put or conveyor approaches on the Frontier supercomputer. Our preliminary results show significant improvements in both performance and energy consumption. These findings suggest that message aggregation can play an important role in addressing the challenges of PGAS energy consumption in modern HPC systems.

KW - energy efficiency

KW - message aggregation

KW - OpenSHMEM

KW - PGAS programming models

UR - https://www.scopus.com/pages/publications/105010334811

U2 - 10.1007/978-3-031-97492-2_12

DO - 10.1007/978-3-031-97492-2_12

M3 - Chapter

AN - SCOPUS:105010334811

T3 - Lecture Notes in Computer Science

SP - 114

EP - 120

BT - Lecture Notes in Computer Science

PB - Springer Science and Business Media Deutschland GmbH

ER -