A holistic approach towards intelligent hotspot prevention in network-on-chip-based multicores

Vassos Soteriou Soteriou, Theocharis Theocharides, Elena Kakoulli

Research output: Contribution to journalArticle

Abstract

Traffic hotspots, a severe form of network congestion, can be caused unexpectedly in a network-on-chip (NoC) due to the immanent spatio-temporal unevenness of application traffic. Hotspots reduce the NoC's effective throughput, where in the worst-case scenario, network traffic flows can be frozen indefinitely. To alleviate this problematic phenomenon several adaptive routing algorithms employ online load-balancing schemes, aiming to reduce the possibility of hotspots arising. Since most are not explicitly hotspot-agnostic, they cannot completely prevent hotspot formation(s) as their reactive capability to hotspots is merely passive. This paper presents a pro-active Hotspot-Preventive Routing Algorithm (HPRA) which uses the advance knowledge gained from network-embedded artificial neural network-based (ANN) hotspot predictors to guide packet routing in mitigating any unforeseen near-future hotspot occurrences. First, these ANN-based predictors are trained offline and during multicore operation they gather online statistical data to predict about-to-be-formed hotspots, promptly informing HPRA to take appropriate hotspot-preventive action(s). Next, in a holistic approach, additional ANN training is performed with data acquired after HPRA interferes, so as to further improve hotspot prediction accuracy; hence, the ANN mechanism does not only predict hotspots, but is also aware of changes that HPRA imposes upon the interconnect infrastructure. Evaluation results, including utilizing real application traffic traces gathered from parallelized workload executions onto a chip multiprocessor architecture, show that HPRA can improve network throughput up to 81 percent when compared with prior-art. Hardware synthesis results affirm the HPRA mechanism's moderate overhead requisites.

Original languageEnglish (US)
Article number7110592
Pages (from-to)819-833
Number of pages15
JournalIEEE Transactions on Computers
Volume65
Issue number3
DOIs
StatePublished - Mar 1 2016

Fingerprint

Routing algorithms
Hot Spot
Routing Algorithm
Neural networks
Throughput
Artificial Neural Network
Adaptive algorithms
Network on chip
Network-on-chip
Resource allocation
Traffic
Predictors
Hardware
Packet Routing
Adaptive Routing
Predict
Chip multiprocessors
Network Flow
Network Traffic
Interconnect

Keywords

  • Multiprocessor interconnection
  • neural network hardware
  • on-chip network
  • ultra-scale integration

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Software
  • Hardware and Architecture
  • Computational Theory and Mathematics

Cite this

A holistic approach towards intelligent hotspot prevention in network-on-chip-based multicores. / Soteriou, Vassos Soteriou; Theocharides, Theocharis; Kakoulli, Elena.

In: IEEE Transactions on Computers, Vol. 65, No. 3, 7110592, 01.03.2016, p. 819-833.

Research output: Contribution to journalArticle

@article{dea7ede9ede349258f6b860d9375090d,
title = "A holistic approach towards intelligent hotspot prevention in network-on-chip-based multicores",
abstract = "Traffic hotspots, a severe form of network congestion, can be caused unexpectedly in a network-on-chip (NoC) due to the immanent spatio-temporal unevenness of application traffic. Hotspots reduce the NoC's effective throughput, where in the worst-case scenario, network traffic flows can be frozen indefinitely. To alleviate this problematic phenomenon several adaptive routing algorithms employ online load-balancing schemes, aiming to reduce the possibility of hotspots arising. Since most are not explicitly hotspot-agnostic, they cannot completely prevent hotspot formation(s) as their reactive capability to hotspots is merely passive. This paper presents a pro-active Hotspot-Preventive Routing Algorithm (HPRA) which uses the advance knowledge gained from network-embedded artificial neural network-based (ANN) hotspot predictors to guide packet routing in mitigating any unforeseen near-future hotspot occurrences. First, these ANN-based predictors are trained offline and during multicore operation they gather online statistical data to predict about-to-be-formed hotspots, promptly informing HPRA to take appropriate hotspot-preventive action(s). Next, in a holistic approach, additional ANN training is performed with data acquired after HPRA interferes, so as to further improve hotspot prediction accuracy; hence, the ANN mechanism does not only predict hotspots, but is also aware of changes that HPRA imposes upon the interconnect infrastructure. Evaluation results, including utilizing real application traffic traces gathered from parallelized workload executions onto a chip multiprocessor architecture, show that HPRA can improve network throughput up to 81 percent when compared with prior-art. Hardware synthesis results affirm the HPRA mechanism's moderate overhead requisites.",
keywords = "Multiprocessor interconnection, neural network hardware, on-chip network, ultra-scale integration",
author = "Soteriou, {Vassos Soteriou} and Theocharis Theocharides and Elena Kakoulli",
year = "2016",
month = "3",
day = "1",
doi = "10.1109/TC.2015.2435748",
language = "English (US)",
volume = "65",
pages = "819--833",
journal = "IEEE Transactions on Computers",
issn = "0018-9340",
publisher = "IEEE Computer Society",
number = "3",

}

TY - JOUR

T1 - A holistic approach towards intelligent hotspot prevention in network-on-chip-based multicores

AU - Soteriou, Vassos Soteriou

AU - Theocharides, Theocharis

AU - Kakoulli, Elena

PY - 2016/3/1

Y1 - 2016/3/1

N2 - Traffic hotspots, a severe form of network congestion, can be caused unexpectedly in a network-on-chip (NoC) due to the immanent spatio-temporal unevenness of application traffic. Hotspots reduce the NoC's effective throughput, where in the worst-case scenario, network traffic flows can be frozen indefinitely. To alleviate this problematic phenomenon several adaptive routing algorithms employ online load-balancing schemes, aiming to reduce the possibility of hotspots arising. Since most are not explicitly hotspot-agnostic, they cannot completely prevent hotspot formation(s) as their reactive capability to hotspots is merely passive. This paper presents a pro-active Hotspot-Preventive Routing Algorithm (HPRA) which uses the advance knowledge gained from network-embedded artificial neural network-based (ANN) hotspot predictors to guide packet routing in mitigating any unforeseen near-future hotspot occurrences. First, these ANN-based predictors are trained offline and during multicore operation they gather online statistical data to predict about-to-be-formed hotspots, promptly informing HPRA to take appropriate hotspot-preventive action(s). Next, in a holistic approach, additional ANN training is performed with data acquired after HPRA interferes, so as to further improve hotspot prediction accuracy; hence, the ANN mechanism does not only predict hotspots, but is also aware of changes that HPRA imposes upon the interconnect infrastructure. Evaluation results, including utilizing real application traffic traces gathered from parallelized workload executions onto a chip multiprocessor architecture, show that HPRA can improve network throughput up to 81 percent when compared with prior-art. Hardware synthesis results affirm the HPRA mechanism's moderate overhead requisites.

AB - Traffic hotspots, a severe form of network congestion, can be caused unexpectedly in a network-on-chip (NoC) due to the immanent spatio-temporal unevenness of application traffic. Hotspots reduce the NoC's effective throughput, where in the worst-case scenario, network traffic flows can be frozen indefinitely. To alleviate this problematic phenomenon several adaptive routing algorithms employ online load-balancing schemes, aiming to reduce the possibility of hotspots arising. Since most are not explicitly hotspot-agnostic, they cannot completely prevent hotspot formation(s) as their reactive capability to hotspots is merely passive. This paper presents a pro-active Hotspot-Preventive Routing Algorithm (HPRA) which uses the advance knowledge gained from network-embedded artificial neural network-based (ANN) hotspot predictors to guide packet routing in mitigating any unforeseen near-future hotspot occurrences. First, these ANN-based predictors are trained offline and during multicore operation they gather online statistical data to predict about-to-be-formed hotspots, promptly informing HPRA to take appropriate hotspot-preventive action(s). Next, in a holistic approach, additional ANN training is performed with data acquired after HPRA interferes, so as to further improve hotspot prediction accuracy; hence, the ANN mechanism does not only predict hotspots, but is also aware of changes that HPRA imposes upon the interconnect infrastructure. Evaluation results, including utilizing real application traffic traces gathered from parallelized workload executions onto a chip multiprocessor architecture, show that HPRA can improve network throughput up to 81 percent when compared with prior-art. Hardware synthesis results affirm the HPRA mechanism's moderate overhead requisites.

KW - Multiprocessor interconnection

KW - neural network hardware

KW - on-chip network

KW - ultra-scale integration

UR - http://www.scopus.com/inward/record.url?scp=84962127868&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84962127868&partnerID=8YFLogxK

U2 - 10.1109/TC.2015.2435748

DO - 10.1109/TC.2015.2435748

M3 - Article

AN - SCOPUS:84962127868

VL - 65

SP - 819

EP - 833

JO - IEEE Transactions on Computers

JF - IEEE Transactions on Computers

SN - 0018-9340

IS - 3

M1 - 7110592

ER -