Predicting Local Congestion at Fine-grain Levels in Networks-on-Chip Using Spiking Neural Networks

Aqib Javed; Jim Harkin; LJ McDaid; Junxiu Liu

Predicting Local Congestion at Fine-grain Levels in Networks-on-Chip Using Spiking Neural Networks

Aqib Javed, Jim Harkin, LJ McDaid, Junxiu Liu

Research output: Contribution to conference › Paper › peer-review

98 Downloads (Pure)

Abstract

Networks-on-Chip (NoC) was introduced to achieve
maximum communication performance in on-chip systems. Local
congestion caused by the queuing of data at input channel buffers
constrains NoC latency and throughput performance. NoCs
require a predictive approach to minimize the effects from local
congestion problems. In this paper we proposed a novel fine-grain
congestion prediction approach based on Spiking Neural Network
(SNN), which predicts router congestion with 30 clock cycles look-
ahead capability. Two fine-grain prediction approaches including
router and network models are proposed. The prediction
performances of the models are evaluated and analyzed using both
synthetic and real-time NoC traffic applications. Results show that
the network model is more consistent in fine-grain local congestion
prediction and requires 42% less hardware area than the router
model.

Original language	English
Pages	1-7
Number of pages	7
Publication status	Accepted/In press - 15 Sept 2020
Event	13th International Workshop on Network on Chip Architectures - Online Duration: 18 Oct 2020 → … http://www.nocarc.org/home

Workshop

Workshop	13th International Workshop on Network on Chip Architectures
Period	18/10/20 → …
Internet address	http://www.nocarc.org/home

Keywords

Networks-on-chip
congestion prediction
network traffic
Spiking Neural Networks

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

Accepted ManuscriptAuthor Accepted Manuscript, 1.82 MB

Cite this

@conference{df81cd93d02440bd940bd8b65734f766,

title = "Predicting Local Congestion at Fine-grain Levels in Networks-on-Chip Using Spiking Neural Networks",

abstract = "Networks-on-Chip (NoC) was introduced to achieve maximum communication performance in on-chip systems. Local congestion caused by the queuing of data at input channel buffers constrains NoC latency and throughput performance. NoCs require a predictive approach to minimize the effects from local congestion problems. In this paper we proposed a novel fine-grain congestion prediction approach based on Spiking Neural Network (SNN), which predicts router congestion with 30 clock cycles look-ahead capability. Two fine-grain prediction approaches including router and network models are proposed. The prediction performances of the models are evaluated and analyzed using both synthetic and real-time NoC traffic applications. Results show that the network model is more consistent in fine-grain local congestion prediction and requires 42% less hardware area than the router model. ",

keywords = "Networks-on-chip, congestion prediction, network traffic, Spiking Neural Networks",

author = "Aqib Javed and Jim Harkin and LJ McDaid and Junxiu Liu",

year = "2020",

month = sep,

day = "15",

language = "English",

pages = "1--7",

note = "13th International Workshop on Network on Chip Architectures ; Conference date: 18-10-2020",

url = "http://www.nocarc.org/home",

}

TY - CONF

T1 - Predicting Local Congestion at Fine-grain Levels in Networks-on-Chip Using Spiking Neural Networks

AU - Javed, Aqib

AU - Harkin, Jim

AU - McDaid, LJ

AU - Liu, Junxiu

PY - 2020/9/15

Y1 - 2020/9/15

N2 - Networks-on-Chip (NoC) was introduced to achieve maximum communication performance in on-chip systems. Local congestion caused by the queuing of data at input channel buffers constrains NoC latency and throughput performance. NoCs require a predictive approach to minimize the effects from local congestion problems. In this paper we proposed a novel fine-grain congestion prediction approach based on Spiking Neural Network (SNN), which predicts router congestion with 30 clock cycles look-ahead capability. Two fine-grain prediction approaches including router and network models are proposed. The prediction performances of the models are evaluated and analyzed using both synthetic and real-time NoC traffic applications. Results show that the network model is more consistent in fine-grain local congestion prediction and requires 42% less hardware area than the router model.

AB - Networks-on-Chip (NoC) was introduced to achieve maximum communication performance in on-chip systems. Local congestion caused by the queuing of data at input channel buffers constrains NoC latency and throughput performance. NoCs require a predictive approach to minimize the effects from local congestion problems. In this paper we proposed a novel fine-grain congestion prediction approach based on Spiking Neural Network (SNN), which predicts router congestion with 30 clock cycles look-ahead capability. Two fine-grain prediction approaches including router and network models are proposed. The prediction performances of the models are evaluated and analyzed using both synthetic and real-time NoC traffic applications. Results show that the network model is more consistent in fine-grain local congestion prediction and requires 42% less hardware area than the router model.

KW - Networks-on-chip

KW - congestion prediction

KW - network traffic

KW - Spiking Neural Networks

M3 - Paper

SP - 1

EP - 7

T2 - 13th International Workshop on Network on Chip Architectures

Y2 - 18 October 2020

ER -

Predicting Local Congestion at Fine-grain Levels in Networks-on-Chip Using Spiking Neural Networks

Abstract

Workshop

Keywords

UN SDGs

Access to Document

Fingerprint

Cite this