Inter-node Message Passing through Optical Reconfigurable Memory Channel

Mauricio G. Palma; Jorge Gonzalez; Martin Carrasco; Ruth Rubio-Noriega; Keren Bergman; Rodolfo Azevedo

doi:10.1109/ACCESS.2024.3412878

Inter-node Message Passing through Optical Reconfigurable Memory Channel

Mauricio G. Palma, Jorge Gonzalez, Martin Carrasco, Ruth Rubio-Noriega, Keren Bergman, Rodolfo Azevedo

Department of Computer Science

Research output: Contribution to journal › Article › peer-review

1 Scopus citations

Abstract

Efficient data movement between nodes in a data center is essential for optimal performance of distributed workloads. With advancements in computing interconnection and memory, new opportunities have emerged. We propose a novel inter-node architecture and protocol called Flexible Memory Units (FMU) that uses optically disaggregated memory. FMUs can be dynamically allocated to different nodes during runtime using optical switches. The primary objective of FMUs is to use the disaggregated memory as temporary buffers during inter-node communication. We have implemented Simplecomm, an open-source simulator, to evaluate real MPI benchmarks using FMU. Our evaluation demonstrates significant speedups of up to 5.18× in communication-bound applications and 1.22× on computing-intensive applications, compared to a 100 Gbps InfiniBand interconnect.

Original language	English
Pages (from-to)	1
Number of pages	1
Journal	IEEE Access
DOIs	https://doi.org/10.1109/ACCESS.2024.3412878
State	Accepted/In press - 2024

Keywords

Computer architecture
disaggregated memory
Inter-node message passing
Memory management
message passing
MPI
Optical buffering
optical interconnection
Optical interconnections
Optical switches
photonic
Photonics
Protocols

Access to Document

10.1109/ACCESS.2024.3412878

Cite this

@article{601220254aff4743b4667ec853976867,

title = "Inter-node Message Passing through Optical Reconfigurable Memory Channel",

abstract = "Efficient data movement between nodes in a data center is essential for optimal performance of distributed workloads. With advancements in computing interconnection and memory, new opportunities have emerged. We propose a novel inter-node architecture and protocol called Flexible Memory Units (FMU) that uses optically disaggregated memory. FMUs can be dynamically allocated to different nodes during runtime using optical switches. The primary objective of FMUs is to use the disaggregated memory as temporary buffers during inter-node communication. We have implemented Simplecomm, an open-source simulator, to evaluate real MPI benchmarks using FMU. Our evaluation demonstrates significant speedups of up to 5.18× in communication-bound applications and 1.22× on computing-intensive applications, compared to a 100 Gbps InfiniBand interconnect.",

keywords = "Computer architecture, disaggregated memory, Inter-node message passing, Memory management, message passing, MPI, Optical buffering, optical interconnection, Optical interconnections, Optical switches, photonic, Photonics, Protocols",

author = "Palma, {Mauricio G.} and Jorge Gonzalez and Martin Carrasco and Ruth Rubio-Noriega and Keren Bergman and Rodolfo Azevedo",

note = "Publisher Copyright: Authors",

year = "2024",

doi = "10.1109/ACCESS.2024.3412878",

language = "English",

pages = "1",

journal = "IEEE Access",

issn = "2169-3536",

}

TY - JOUR

T1 - Inter-node Message Passing through Optical Reconfigurable Memory Channel

AU - Palma, Mauricio G.

AU - Gonzalez, Jorge

AU - Carrasco, Martin

AU - Rubio-Noriega, Ruth

AU - Bergman, Keren

AU - Azevedo, Rodolfo

N1 - Publisher Copyright: Authors

PY - 2024

Y1 - 2024

N2 - Efficient data movement between nodes in a data center is essential for optimal performance of distributed workloads. With advancements in computing interconnection and memory, new opportunities have emerged. We propose a novel inter-node architecture and protocol called Flexible Memory Units (FMU) that uses optically disaggregated memory. FMUs can be dynamically allocated to different nodes during runtime using optical switches. The primary objective of FMUs is to use the disaggregated memory as temporary buffers during inter-node communication. We have implemented Simplecomm, an open-source simulator, to evaluate real MPI benchmarks using FMU. Our evaluation demonstrates significant speedups of up to 5.18× in communication-bound applications and 1.22× on computing-intensive applications, compared to a 100 Gbps InfiniBand interconnect.

AB - Efficient data movement between nodes in a data center is essential for optimal performance of distributed workloads. With advancements in computing interconnection and memory, new opportunities have emerged. We propose a novel inter-node architecture and protocol called Flexible Memory Units (FMU) that uses optically disaggregated memory. FMUs can be dynamically allocated to different nodes during runtime using optical switches. The primary objective of FMUs is to use the disaggregated memory as temporary buffers during inter-node communication. We have implemented Simplecomm, an open-source simulator, to evaluate real MPI benchmarks using FMU. Our evaluation demonstrates significant speedups of up to 5.18× in communication-bound applications and 1.22× on computing-intensive applications, compared to a 100 Gbps InfiniBand interconnect.

KW - Computer architecture

KW - disaggregated memory

KW - Inter-node message passing

KW - Memory management

KW - message passing

KW - MPI

KW - Optical buffering

KW - optical interconnection

KW - Optical interconnections

KW - Optical switches

KW - photonic

KW - Photonics

KW - Protocols

UR - http://www.scopus.com/inward/record.url?scp=85196109251&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2024.3412878

DO - 10.1109/ACCESS.2024.3412878

M3 - Article

AN - SCOPUS:85196109251

SN - 2169-3536

SP - 1

JO - IEEE Access

JF - IEEE Access

ER -

Inter-node Message Passing through Optical Reconfigurable Memory Channel

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this