Design of a high-speed optical interconnect for scalable shared memory multiprocessors

Avinash Karanth Kodi, Ahmed Louri

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

4 Citations (Scopus)

Abstract

This paper proposes a highly connected optical interconnect based architecture that maximizes the channel availability for future scalable parallel computers such as Distributed Shared Memory (DSM) multiprocessors and cluster networks. As the system size increases, various messages (requests, responses and acknowledgments) increase in the network resulting in contention. This results in increasing the remote memory access latency and significantly affects the performance of these parallel computers. As a solution, we propose an architecture called RAPID (Reconfigurable and scalable All-Photonic Interconnect for Distributed-shared memory), that provides low remote memory access latency by providing fast and efficient unicast, multicast and broadcast capabilities using a combination of aggressively designed WDM, TDM and SDM techniques. We evaluated RAPID based on network characteristics and by simulation using synthetic traffic workloads and compared it against other networks such as electrical ring, torus, mesh and hypercube networks. We found that RAPID outperforms all networks and satisfies most of the requirements of parallel computer design such as low latency, high bandwidth, high connectivity, and easy scalability.

Original languageEnglish (US)
Title of host publicationProceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects
EditorsS. Watters
Pages92-97
Number of pages6
DOIs
StatePublished - 2004
EventProceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects - Stanford, CA, United States
Duration: Aug 25 2004Aug 27 2004

Other

OtherProceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects
CountryUnited States
CityStanford, CA
Period8/25/048/27/04

Fingerprint

Optical interconnects
Data storage equipment
Hypercube networks
Time division multiplexing
Wavelength division multiplexing
Photonics
Scalability
Availability
Bandwidth

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Kodi, A. K., & Louri, A. (2004). Design of a high-speed optical interconnect for scalable shared memory multiprocessors. In S. Watters (Ed.), Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects (pp. 92-97) https://doi.org/10.1109/CONECT.2004.1375210

Design of a high-speed optical interconnect for scalable shared memory multiprocessors. / Kodi, Avinash Karanth; Louri, Ahmed.

Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects. ed. / S. Watters. 2004. p. 92-97.

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

Kodi, AK & Louri, A 2004, Design of a high-speed optical interconnect for scalable shared memory multiprocessors. in S Watters (ed.), Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects. pp. 92-97, Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects, Stanford, CA, United States, 8/25/04. https://doi.org/10.1109/CONECT.2004.1375210
Kodi AK, Louri A. Design of a high-speed optical interconnect for scalable shared memory multiprocessors. In Watters S, editor, Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects. 2004. p. 92-97 https://doi.org/10.1109/CONECT.2004.1375210
Kodi, Avinash Karanth ; Louri, Ahmed. / Design of a high-speed optical interconnect for scalable shared memory multiprocessors. Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects. editor / S. Watters. 2004. pp. 92-97
@inproceedings{9381f5ff1d334e2ab885463809489c0d,
title = "Design of a high-speed optical interconnect for scalable shared memory multiprocessors",
abstract = "This paper proposes a highly connected optical interconnect based architecture that maximizes the channel availability for future scalable parallel computers such as Distributed Shared Memory (DSM) multiprocessors and cluster networks. As the system size increases, various messages (requests, responses and acknowledgments) increase in the network resulting in contention. This results in increasing the remote memory access latency and significantly affects the performance of these parallel computers. As a solution, we propose an architecture called RAPID (Reconfigurable and scalable All-Photonic Interconnect for Distributed-shared memory), that provides low remote memory access latency by providing fast and efficient unicast, multicast and broadcast capabilities using a combination of aggressively designed WDM, TDM and SDM techniques. We evaluated RAPID based on network characteristics and by simulation using synthetic traffic workloads and compared it against other networks such as electrical ring, torus, mesh and hypercube networks. We found that RAPID outperforms all networks and satisfies most of the requirements of parallel computer design such as low latency, high bandwidth, high connectivity, and easy scalability.",
author = "Kodi, {Avinash Karanth} and Ahmed Louri",
year = "2004",
doi = "10.1109/CONECT.2004.1375210",
language = "English (US)",
isbn = "0780386868",
pages = "92--97",
editor = "S. Watters",
booktitle = "Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects",

}

TY - GEN

T1 - Design of a high-speed optical interconnect for scalable shared memory multiprocessors

AU - Kodi, Avinash Karanth

AU - Louri, Ahmed

PY - 2004

Y1 - 2004

N2 - This paper proposes a highly connected optical interconnect based architecture that maximizes the channel availability for future scalable parallel computers such as Distributed Shared Memory (DSM) multiprocessors and cluster networks. As the system size increases, various messages (requests, responses and acknowledgments) increase in the network resulting in contention. This results in increasing the remote memory access latency and significantly affects the performance of these parallel computers. As a solution, we propose an architecture called RAPID (Reconfigurable and scalable All-Photonic Interconnect for Distributed-shared memory), that provides low remote memory access latency by providing fast and efficient unicast, multicast and broadcast capabilities using a combination of aggressively designed WDM, TDM and SDM techniques. We evaluated RAPID based on network characteristics and by simulation using synthetic traffic workloads and compared it against other networks such as electrical ring, torus, mesh and hypercube networks. We found that RAPID outperforms all networks and satisfies most of the requirements of parallel computer design such as low latency, high bandwidth, high connectivity, and easy scalability.

AB - This paper proposes a highly connected optical interconnect based architecture that maximizes the channel availability for future scalable parallel computers such as Distributed Shared Memory (DSM) multiprocessors and cluster networks. As the system size increases, various messages (requests, responses and acknowledgments) increase in the network resulting in contention. This results in increasing the remote memory access latency and significantly affects the performance of these parallel computers. As a solution, we propose an architecture called RAPID (Reconfigurable and scalable All-Photonic Interconnect for Distributed-shared memory), that provides low remote memory access latency by providing fast and efficient unicast, multicast and broadcast capabilities using a combination of aggressively designed WDM, TDM and SDM techniques. We evaluated RAPID based on network characteristics and by simulation using synthetic traffic workloads and compared it against other networks such as electrical ring, torus, mesh and hypercube networks. We found that RAPID outperforms all networks and satisfies most of the requirements of parallel computer design such as low latency, high bandwidth, high connectivity, and easy scalability.

UR - http://www.scopus.com/inward/record.url?scp=14844329044&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=14844329044&partnerID=8YFLogxK

U2 - 10.1109/CONECT.2004.1375210

DO - 10.1109/CONECT.2004.1375210

M3 - Conference contribution

SN - 0780386868

SN - 9780780386860

SP - 92

EP - 97

BT - Proceedings - 12th Annual IEEE Symposium on High Performance Interconnects, Hot Interconnects

A2 - Watters, S.

ER -