Intelligent NOC hotspot prediction

Elena Kakoulli, Vassos Soteriou Soteriou, Theocharis Theocharides

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

Hotspots are Network on-Chip (NoC) routers or modules which occasionally receive packetized traffic at a higher rate that they can process. This phenomenon reduces the performance of an NoC, especially in the case wormhole flow-control. Such situations may also lead to deadlocks, raising the need of a hotspot prevention mechanism. Such mechanism can potentially enable the system to adjust its behavior and prevent hotspot formation, subsequently sustaining performance and efficiency. This Chapter presents an Artificial Neural Network-based (ANN) hotspot prediction mechanism, potentially triggering a hotspot avoidance mechanism before the hotspot is formed. The ANN monitors buffer utilization and reactively predicts the location of an about to-be-formed hotspot, allowing enough time for the system to react to these potential hotspots. The neural network is trained using synthetic traffic models, and evaluated using both synthetic and real application traces. Results indicate that a relatively small neural network can predict hotspot formation with accuracy ranges between 76 and 92%.

Original languageEnglish (US)
Title of host publicationVLSI 2010 Annual Symposium
Subtitle of host publicationSelected papers
EditorsNikolaos Voros, Amar Mukherjee, Nicolas Sklavos, Konstantinos Masselos, Michael Huebner
Pages3-16
Number of pages14
DOIs
StatePublished - Dec 1 2011

Publication series

NameLecture Notes in Electrical Engineering
Volume105 LNEE
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Fingerprint

Neural networks
Routers
Flow control
Network-on-chip

Keywords

  • Artificial Neural Networks
  • Network on-Chip Hotspots
  • VLSI Systems

ASJC Scopus subject areas

  • Industrial and Manufacturing Engineering

Cite this

Kakoulli, E., Soteriou, V. S., & Theocharides, T. (2011). Intelligent NOC hotspot prediction. In N. Voros, A. Mukherjee, N. Sklavos, K. Masselos, & M. Huebner (Eds.), VLSI 2010 Annual Symposium: Selected papers (pp. 3-16). (Lecture Notes in Electrical Engineering; Vol. 105 LNEE). https://doi.org/10.1007/978-94-007-1488-5_1

Intelligent NOC hotspot prediction. / Kakoulli, Elena; Soteriou, Vassos Soteriou; Theocharides, Theocharis.

VLSI 2010 Annual Symposium: Selected papers. ed. / Nikolaos Voros; Amar Mukherjee; Nicolas Sklavos; Konstantinos Masselos; Michael Huebner. 2011. p. 3-16 (Lecture Notes in Electrical Engineering; Vol. 105 LNEE).

Research output: Chapter in Book/Report/Conference proceedingChapter

Kakoulli, E, Soteriou, VS & Theocharides, T 2011, Intelligent NOC hotspot prediction. in N Voros, A Mukherjee, N Sklavos, K Masselos & M Huebner (eds), VLSI 2010 Annual Symposium: Selected papers. Lecture Notes in Electrical Engineering, vol. 105 LNEE, pp. 3-16. https://doi.org/10.1007/978-94-007-1488-5_1
Kakoulli E, Soteriou VS, Theocharides T. Intelligent NOC hotspot prediction. In Voros N, Mukherjee A, Sklavos N, Masselos K, Huebner M, editors, VLSI 2010 Annual Symposium: Selected papers. 2011. p. 3-16. (Lecture Notes in Electrical Engineering). https://doi.org/10.1007/978-94-007-1488-5_1
Kakoulli, Elena ; Soteriou, Vassos Soteriou ; Theocharides, Theocharis. / Intelligent NOC hotspot prediction. VLSI 2010 Annual Symposium: Selected papers. editor / Nikolaos Voros ; Amar Mukherjee ; Nicolas Sklavos ; Konstantinos Masselos ; Michael Huebner. 2011. pp. 3-16 (Lecture Notes in Electrical Engineering).
@inbook{8a09982759c846c795b6e5e1adbe41f9,
title = "Intelligent NOC hotspot prediction",
abstract = "Hotspots are Network on-Chip (NoC) routers or modules which occasionally receive packetized traffic at a higher rate that they can process. This phenomenon reduces the performance of an NoC, especially in the case wormhole flow-control. Such situations may also lead to deadlocks, raising the need of a hotspot prevention mechanism. Such mechanism can potentially enable the system to adjust its behavior and prevent hotspot formation, subsequently sustaining performance and efficiency. This Chapter presents an Artificial Neural Network-based (ANN) hotspot prediction mechanism, potentially triggering a hotspot avoidance mechanism before the hotspot is formed. The ANN monitors buffer utilization and reactively predicts the location of an about to-be-formed hotspot, allowing enough time for the system to react to these potential hotspots. The neural network is trained using synthetic traffic models, and evaluated using both synthetic and real application traces. Results indicate that a relatively small neural network can predict hotspot formation with accuracy ranges between 76 and 92{\%}.",
keywords = "Artificial Neural Networks, Network on-Chip Hotspots, VLSI Systems",
author = "Elena Kakoulli and Soteriou, {Vassos Soteriou} and Theocharis Theocharides",
year = "2011",
month = "12",
day = "1",
doi = "10.1007/978-94-007-1488-5_1",
language = "English (US)",
isbn = "9789400714878",
series = "Lecture Notes in Electrical Engineering",
pages = "3--16",
editor = "Nikolaos Voros and Amar Mukherjee and Nicolas Sklavos and Konstantinos Masselos and Michael Huebner",
booktitle = "VLSI 2010 Annual Symposium",

}

TY - CHAP

T1 - Intelligent NOC hotspot prediction

AU - Kakoulli, Elena

AU - Soteriou, Vassos Soteriou

AU - Theocharides, Theocharis

PY - 2011/12/1

Y1 - 2011/12/1

N2 - Hotspots are Network on-Chip (NoC) routers or modules which occasionally receive packetized traffic at a higher rate that they can process. This phenomenon reduces the performance of an NoC, especially in the case wormhole flow-control. Such situations may also lead to deadlocks, raising the need of a hotspot prevention mechanism. Such mechanism can potentially enable the system to adjust its behavior and prevent hotspot formation, subsequently sustaining performance and efficiency. This Chapter presents an Artificial Neural Network-based (ANN) hotspot prediction mechanism, potentially triggering a hotspot avoidance mechanism before the hotspot is formed. The ANN monitors buffer utilization and reactively predicts the location of an about to-be-formed hotspot, allowing enough time for the system to react to these potential hotspots. The neural network is trained using synthetic traffic models, and evaluated using both synthetic and real application traces. Results indicate that a relatively small neural network can predict hotspot formation with accuracy ranges between 76 and 92%.

AB - Hotspots are Network on-Chip (NoC) routers or modules which occasionally receive packetized traffic at a higher rate that they can process. This phenomenon reduces the performance of an NoC, especially in the case wormhole flow-control. Such situations may also lead to deadlocks, raising the need of a hotspot prevention mechanism. Such mechanism can potentially enable the system to adjust its behavior and prevent hotspot formation, subsequently sustaining performance and efficiency. This Chapter presents an Artificial Neural Network-based (ANN) hotspot prediction mechanism, potentially triggering a hotspot avoidance mechanism before the hotspot is formed. The ANN monitors buffer utilization and reactively predicts the location of an about to-be-formed hotspot, allowing enough time for the system to react to these potential hotspots. The neural network is trained using synthetic traffic models, and evaluated using both synthetic and real application traces. Results indicate that a relatively small neural network can predict hotspot formation with accuracy ranges between 76 and 92%.

KW - Artificial Neural Networks

KW - Network on-Chip Hotspots

KW - VLSI Systems

UR - http://www.scopus.com/inward/record.url?scp=84856592059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84856592059&partnerID=8YFLogxK

U2 - 10.1007/978-94-007-1488-5_1

DO - 10.1007/978-94-007-1488-5_1

M3 - Chapter

AN - SCOPUS:84856592059

SN - 9789400714878

T3 - Lecture Notes in Electrical Engineering

SP - 3

EP - 16

BT - VLSI 2010 Annual Symposium

A2 - Voros, Nikolaos

A2 - Mukherjee, Amar

A2 - Sklavos, Nicolas

A2 - Masselos, Konstantinos

A2 - Huebner, Michael

ER -