An artificial neural network-based hotspot prediction mechanism for NoCs

Elena Kakoullit, Vassos Soteriou Soteriou, Theocharis Theocharides

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

Abstract

Hotspots are network on-chip (NoC) routers or modules in systems on-chip (SoCs) which occasionally receive packetized traffic at a rate higher than they can consume It. This adverse phenomenon greatly reduces the performance of an NoC, especially in the case of today's widely-employed wormhole flow-control, as backpressure can cause the buffers of neighboring routers to quickly fill-up leading to a spatial spread in congestion that can cause the network to saturate. Even worse, such situations may lead to deadlocks. Thus, a hotspot prevention mechanism can be greatly beneficial, as it can potentially enable the interconnection system to adjust its behavior and prevent the rise of potential hotspots, subsequently sustaining NoC performance and efficiency. Unfortunately, hotspots cannot be known a-priori In NoCs used in general-purpose systems as application demands are not predetermined unlike in application-specific SoCs, making hotspot prediction and subsequently prevention difficult. In this paper we present an artificial neural network-based hotspot prediction mechanism that can be potentially used in tandem with a hotspot avoidance mechanism for handling an unforeseen hotspot formation efficiently. The network uses buffer utilization statistical data to dynamically monitor the interconnect fabric, 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% to 92% when evaluated on two different mesh NoCs.

Original languageEnglish (US)
Title of host publicationProceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010
Pages339-344
Number of pages6
DOIs
StatePublished - Oct 20 2010
EventIEEE Annual Symposium on VLSI, ISVLSI 2010 - Lixouri, Kefalonia, Greece
Duration: Jul 5 2010Jul 7 2010

Other

OtherIEEE Annual Symposium on VLSI, ISVLSI 2010
CountryGreece
CityLixouri, Kefalonia
Period7/5/107/7/10

Fingerprint

Neural networks
Routers
Flow control
Network-on-chip
System-on-chip

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this

Kakoullit, E., Soteriou, V. S., & Theocharides, T. (2010). An artificial neural network-based hotspot prediction mechanism for NoCs. In Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010 (pp. 339-344). [5572797] https://doi.org/10.1109/ISVLSI.2010.50

An artificial neural network-based hotspot prediction mechanism for NoCs. / Kakoullit, Elena; Soteriou, Vassos Soteriou; Theocharides, Theocharis.

Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010. 2010. p. 339-344 5572797.

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

Kakoullit, E, Soteriou, VS & Theocharides, T 2010, An artificial neural network-based hotspot prediction mechanism for NoCs. in Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010., 5572797, pp. 339-344, IEEE Annual Symposium on VLSI, ISVLSI 2010, Lixouri, Kefalonia, Greece, 7/5/10. https://doi.org/10.1109/ISVLSI.2010.50
Kakoullit E, Soteriou VS, Theocharides T. An artificial neural network-based hotspot prediction mechanism for NoCs. In Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010. 2010. p. 339-344. 5572797 https://doi.org/10.1109/ISVLSI.2010.50
Kakoullit, Elena ; Soteriou, Vassos Soteriou ; Theocharides, Theocharis. / An artificial neural network-based hotspot prediction mechanism for NoCs. Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010. 2010. pp. 339-344
@inproceedings{65318b86c8b74c3b894ea0d817251b22,
title = "An artificial neural network-based hotspot prediction mechanism for NoCs",
abstract = "Hotspots are network on-chip (NoC) routers or modules in systems on-chip (SoCs) which occasionally receive packetized traffic at a rate higher than they can consume It. This adverse phenomenon greatly reduces the performance of an NoC, especially in the case of today's widely-employed wormhole flow-control, as backpressure can cause the buffers of neighboring routers to quickly fill-up leading to a spatial spread in congestion that can cause the network to saturate. Even worse, such situations may lead to deadlocks. Thus, a hotspot prevention mechanism can be greatly beneficial, as it can potentially enable the interconnection system to adjust its behavior and prevent the rise of potential hotspots, subsequently sustaining NoC performance and efficiency. Unfortunately, hotspots cannot be known a-priori In NoCs used in general-purpose systems as application demands are not predetermined unlike in application-specific SoCs, making hotspot prediction and subsequently prevention difficult. In this paper we present an artificial neural network-based hotspot prediction mechanism that can be potentially used in tandem with a hotspot avoidance mechanism for handling an unforeseen hotspot formation efficiently. The network uses buffer utilization statistical data to dynamically monitor the interconnect fabric, 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{\%} to 92{\%} when evaluated on two different mesh NoCs.",
author = "Elena Kakoullit and Soteriou, {Vassos Soteriou} and Theocharis Theocharides",
year = "2010",
month = "10",
day = "20",
doi = "10.1109/ISVLSI.2010.50",
language = "English (US)",
isbn = "9780769540764",
pages = "339--344",
booktitle = "Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010",

}

TY - GEN

T1 - An artificial neural network-based hotspot prediction mechanism for NoCs

AU - Kakoullit, Elena

AU - Soteriou, Vassos Soteriou

AU - Theocharides, Theocharis

PY - 2010/10/20

Y1 - 2010/10/20

N2 - Hotspots are network on-chip (NoC) routers or modules in systems on-chip (SoCs) which occasionally receive packetized traffic at a rate higher than they can consume It. This adverse phenomenon greatly reduces the performance of an NoC, especially in the case of today's widely-employed wormhole flow-control, as backpressure can cause the buffers of neighboring routers to quickly fill-up leading to a spatial spread in congestion that can cause the network to saturate. Even worse, such situations may lead to deadlocks. Thus, a hotspot prevention mechanism can be greatly beneficial, as it can potentially enable the interconnection system to adjust its behavior and prevent the rise of potential hotspots, subsequently sustaining NoC performance and efficiency. Unfortunately, hotspots cannot be known a-priori In NoCs used in general-purpose systems as application demands are not predetermined unlike in application-specific SoCs, making hotspot prediction and subsequently prevention difficult. In this paper we present an artificial neural network-based hotspot prediction mechanism that can be potentially used in tandem with a hotspot avoidance mechanism for handling an unforeseen hotspot formation efficiently. The network uses buffer utilization statistical data to dynamically monitor the interconnect fabric, 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% to 92% when evaluated on two different mesh NoCs.

AB - Hotspots are network on-chip (NoC) routers or modules in systems on-chip (SoCs) which occasionally receive packetized traffic at a rate higher than they can consume It. This adverse phenomenon greatly reduces the performance of an NoC, especially in the case of today's widely-employed wormhole flow-control, as backpressure can cause the buffers of neighboring routers to quickly fill-up leading to a spatial spread in congestion that can cause the network to saturate. Even worse, such situations may lead to deadlocks. Thus, a hotspot prevention mechanism can be greatly beneficial, as it can potentially enable the interconnection system to adjust its behavior and prevent the rise of potential hotspots, subsequently sustaining NoC performance and efficiency. Unfortunately, hotspots cannot be known a-priori In NoCs used in general-purpose systems as application demands are not predetermined unlike in application-specific SoCs, making hotspot prediction and subsequently prevention difficult. In this paper we present an artificial neural network-based hotspot prediction mechanism that can be potentially used in tandem with a hotspot avoidance mechanism for handling an unforeseen hotspot formation efficiently. The network uses buffer utilization statistical data to dynamically monitor the interconnect fabric, 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% to 92% when evaluated on two different mesh NoCs.

UR - http://www.scopus.com/inward/record.url?scp=77957902517&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77957902517&partnerID=8YFLogxK

U2 - 10.1109/ISVLSI.2010.50

DO - 10.1109/ISVLSI.2010.50

M3 - Conference contribution

SN - 9780769540764

SP - 339

EP - 344

BT - Proceedings - IEEE Annual Symposium on VLSI, ISVLSI 2010

ER -