Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling

Monobrata Debnath, Dimitris Konstantinou, Chrysostomos Nicopoulos, Giorgos Dimitrakopoulos, Wei Ming Lin, Junghee Lee

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

Abstract

Implementing cost effective congestion control within the Network-on-Chip (NoC) is a major design challenge. When- ever congestion awareness and/or mitigation is desired, ar- chitects typically rely on the use of adaptive routing algo- rithms, which aim to (intelligently) balance the trafic load throughout the NoC. Nevertheless, the hardware cost in- curred by such solutions is quite considerable, since it en- tails the collection/propagation of trafic-related information and the provisioning of deadlock freedom guarantees. In this paper, we explore the potential of simultaneous edge and in-network trafic throttling, as a low-cost alternative to adaptive routing techniques. Without any reliance on adaptivity by the routing algorithm, combined throttling is demonstrated to yield better (in most cases) throughput improvements than state-of-the-art adaptive routing algo- rithms, but at a significantly lower cost.

Original languageEnglish
Title of host publicationProceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017
PublisherAssociation for Computing Machinery
Pages8-11
Number of pages4
ISBN (Electronic)9781450352260
DOIs
Publication statusPublished - 2017 Jan 25
Externally publishedYes
Event2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017 - Stockholm, Sweden
Duration: 2017 Jan 25 → …

Publication series

NameACM International Conference Proceeding Series
VolumePart F128361

Conference

Conference2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017
CountrySweden
CityStockholm
Period17/1/25 → …

Fingerprint

Routing algorithms
Adaptive algorithms
Costs
Throughput
Hardware
Network-on-chip

Keywords

  • Congestion management
  • Low-Cost architecture
  • Multi-core
  • Network-on-chip

ASJC Scopus subject areas

  • Human-Computer Interaction
  • Computer Networks and Communications
  • Computer Vision and Pattern Recognition
  • Software

Cite this

Debnath, M., Konstantinou, D., Nicopoulos, C., Dimitrakopoulos, G., Lin, W. M., & Lee, J. (2017). Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling. In Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017 (pp. 8-11). (ACM International Conference Proceeding Series; Vol. Part F128361). Association for Computing Machinery. https://doi.org/10.1145/3073763.3073764

Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling. / Debnath, Monobrata; Konstantinou, Dimitris; Nicopoulos, Chrysostomos; Dimitrakopoulos, Giorgos; Lin, Wei Ming; Lee, Junghee.

Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017. Association for Computing Machinery, 2017. p. 8-11 (ACM International Conference Proceeding Series; Vol. Part F128361).

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

Debnath, M, Konstantinou, D, Nicopoulos, C, Dimitrakopoulos, G, Lin, WM & Lee, J 2017, Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling. in Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017. ACM International Conference Proceeding Series, vol. Part F128361, Association for Computing Machinery, pp. 8-11, 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017, Stockholm, Sweden, 17/1/25. https://doi.org/10.1145/3073763.3073764
Debnath M, Konstantinou D, Nicopoulos C, Dimitrakopoulos G, Lin WM, Lee J. Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling. In Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017. Association for Computing Machinery. 2017. p. 8-11. (ACM International Conference Proceeding Series). https://doi.org/10.1145/3073763.3073764
Debnath, Monobrata ; Konstantinou, Dimitris ; Nicopoulos, Chrysostomos ; Dimitrakopoulos, Giorgos ; Lin, Wei Ming ; Lee, Junghee. / Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling. Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017. Association for Computing Machinery, 2017. pp. 8-11 (ACM International Conference Proceeding Series).
@inproceedings{4b62ca72514846b8a263c945a340aa5e,
title = "Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling",
abstract = "Implementing cost effective congestion control within the Network-on-Chip (NoC) is a major design challenge. When- ever congestion awareness and/or mitigation is desired, ar- chitects typically rely on the use of adaptive routing algo- rithms, which aim to (intelligently) balance the trafic load throughout the NoC. Nevertheless, the hardware cost in- curred by such solutions is quite considerable, since it en- tails the collection/propagation of trafic-related information and the provisioning of deadlock freedom guarantees. In this paper, we explore the potential of simultaneous edge and in-network trafic throttling, as a low-cost alternative to adaptive routing techniques. Without any reliance on adaptivity by the routing algorithm, combined throttling is demonstrated to yield better (in most cases) throughput improvements than state-of-the-art adaptive routing algo- rithms, but at a significantly lower cost.",
keywords = "Congestion management, Low-Cost architecture, Multi-core, Network-on-chip",
author = "Monobrata Debnath and Dimitris Konstantinou and Chrysostomos Nicopoulos and Giorgos Dimitrakopoulos and Lin, {Wei Ming} and Junghee Lee",
year = "2017",
month = "1",
day = "25",
doi = "10.1145/3073763.3073764",
language = "English",
series = "ACM International Conference Proceeding Series",
publisher = "Association for Computing Machinery",
pages = "8--11",
booktitle = "Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017",

}

TY - GEN

T1 - Low-cost congestion management in networks-on-chip using edge and in-network traffic throttling

AU - Debnath, Monobrata

AU - Konstantinou, Dimitris

AU - Nicopoulos, Chrysostomos

AU - Dimitrakopoulos, Giorgos

AU - Lin, Wei Ming

AU - Lee, Junghee

PY - 2017/1/25

Y1 - 2017/1/25

N2 - Implementing cost effective congestion control within the Network-on-Chip (NoC) is a major design challenge. When- ever congestion awareness and/or mitigation is desired, ar- chitects typically rely on the use of adaptive routing algo- rithms, which aim to (intelligently) balance the trafic load throughout the NoC. Nevertheless, the hardware cost in- curred by such solutions is quite considerable, since it en- tails the collection/propagation of trafic-related information and the provisioning of deadlock freedom guarantees. In this paper, we explore the potential of simultaneous edge and in-network trafic throttling, as a low-cost alternative to adaptive routing techniques. Without any reliance on adaptivity by the routing algorithm, combined throttling is demonstrated to yield better (in most cases) throughput improvements than state-of-the-art adaptive routing algo- rithms, but at a significantly lower cost.

AB - Implementing cost effective congestion control within the Network-on-Chip (NoC) is a major design challenge. When- ever congestion awareness and/or mitigation is desired, ar- chitects typically rely on the use of adaptive routing algo- rithms, which aim to (intelligently) balance the trafic load throughout the NoC. Nevertheless, the hardware cost in- curred by such solutions is quite considerable, since it en- tails the collection/propagation of trafic-related information and the provisioning of deadlock freedom guarantees. In this paper, we explore the potential of simultaneous edge and in-network trafic throttling, as a low-cost alternative to adaptive routing techniques. Without any reliance on adaptivity by the routing algorithm, combined throttling is demonstrated to yield better (in most cases) throughput improvements than state-of-the-art adaptive routing algo- rithms, but at a significantly lower cost.

KW - Congestion management

KW - Low-Cost architecture

KW - Multi-core

KW - Network-on-chip

UR - http://www.scopus.com/inward/record.url?scp=85025124576&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85025124576&partnerID=8YFLogxK

U2 - 10.1145/3073763.3073764

DO - 10.1145/3073763.3073764

M3 - Conference contribution

AN - SCOPUS:85025124576

T3 - ACM International Conference Proceeding Series

SP - 8

EP - 11

BT - Proceedings of the 2nd International Workshop on Advanced Interconnect Solutions and Technologies for Emerging Computing Systems, AISTECS 2017

PB - Association for Computing Machinery

ER -