Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract

Chih Shuo Mei; Chao Wang

doi:10.1145/3450268.3453507

Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract

Chih Shuo Mei
, Chao Wang

Department of Computer Science and Information Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

2 Citations (Scopus)

Abstract

Neural network model training is indispensable for domain-specific Artificial Intelligent Internet-of-Things (AIoT) applications. Typically, a GPU graphics card may take several hundreds watts in average during model training, while an embedded GPU device may take only couple watts for the same purpose at the cost of a longer training time. In this paper, we report our empirical study on the model training using NVIDIA RTX 2080 Ti graphics card and NVIDIA Jetson Nano embedded device. We show that, surprisingly, while the training time using the Jetson Nano is 30 times slower than that using the graphics card, the total energy consumption by Jetson Nano is actually only half. The result suggests that when the response time is less critical, one may choose to do model training on GPU embedded devices instead.

Original language	English
Title of host publication	IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation
Publisher	Association for Computing Machinery, Inc
Pages	253-254
Number of pages	2
ISBN (Electronic)	9781450383547
DOIs	https://doi.org/10.1145/3450268.3453507
Publication status	Published - 2021 May 18
Event	6th ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2021 - Virtual, Online, United States Duration: 2021 May 18 → 2021 May 21

Publication series

Name	IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation

Conference

Conference	6th ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2021
Country/Territory	United States
City	Virtual, Online
Period	2021/05/18 → 2021/05/21

UN SDGs

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

SDG 7 Affordable and Clean Energy

Keywords

Deep Learning
Embedded Systems
Energy Efficiency

ASJC Scopus subject areas

Computer Networks and Communications
Computer Science Applications
Hardware and Architecture

Access to Document

10.1145/3450268.3453507

Cite this

Mei, C. S., & Wang, C. (2021). Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract. In IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation (pp. 253-254). (IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation). Association for Computing Machinery, Inc. https://doi.org/10.1145/3450268.3453507

Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract. / Mei, Chih Shuo; Wang, Chao.
IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation. Association for Computing Machinery, Inc, 2021. p. 253-254 (IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Mei, CS & Wang, C 2021, Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract. in IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation. IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation, Association for Computing Machinery, Inc, pp. 253-254, 6th ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2021, Virtual, Online, United States, 2021/05/18. https://doi.org/10.1145/3450268.3453507

Mei CS, Wang C. Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract. In IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation. Association for Computing Machinery, Inc. 2021. p. 253-254. (IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation). doi: 10.1145/3450268.3453507

Mei, Chih Shuo ; Wang, Chao. / Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications : Poster Abstract. IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation. Association for Computing Machinery, Inc, 2021. pp. 253-254 (IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation).

@inproceedings{2faddbf0c84d422d9b5cba9195b3f205,

title = "Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract",

abstract = "Neural network model training is indispensable for domain-specific Artificial Intelligent Internet-of-Things (AIoT) applications. Typically, a GPU graphics card may take several hundreds watts in average during model training, while an embedded GPU device may take only couple watts for the same purpose at the cost of a longer training time. In this paper, we report our empirical study on the model training using NVIDIA RTX 2080 Ti graphics card and NVIDIA Jetson Nano embedded device. We show that, surprisingly, while the training time using the Jetson Nano is 30 times slower than that using the graphics card, the total energy consumption by Jetson Nano is actually only half. The result suggests that when the response time is less critical, one may choose to do model training on GPU embedded devices instead.",

keywords = "Deep Learning, Embedded Systems, Energy Efficiency",

author = "Mei, \{Chih Shuo\} and Chao Wang",

note = "Publisher Copyright: {\textcopyright} 2021 ACM.; 6th ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2021 ; Conference date: 18-05-2021 Through 21-05-2021",

year = "2021",

month = may,

day = "18",

doi = "10.1145/3450268.3453507",

language = "English",

series = "IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation",

publisher = "Association for Computing Machinery, Inc",

pages = "253--254",

booktitle = "IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation",

}

TY - GEN

T1 - Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications

T2 - 6th ACM/IEEE International Conference on Internet of Things Design and Implementation, IoTDI 2021

AU - Mei, Chih Shuo

AU - Wang, Chao

PY - 2021/5/18

Y1 - 2021/5/18

N2 - Neural network model training is indispensable for domain-specific Artificial Intelligent Internet-of-Things (AIoT) applications. Typically, a GPU graphics card may take several hundreds watts in average during model training, while an embedded GPU device may take only couple watts for the same purpose at the cost of a longer training time. In this paper, we report our empirical study on the model training using NVIDIA RTX 2080 Ti graphics card and NVIDIA Jetson Nano embedded device. We show that, surprisingly, while the training time using the Jetson Nano is 30 times slower than that using the graphics card, the total energy consumption by Jetson Nano is actually only half. The result suggests that when the response time is less critical, one may choose to do model training on GPU embedded devices instead.

AB - Neural network model training is indispensable for domain-specific Artificial Intelligent Internet-of-Things (AIoT) applications. Typically, a GPU graphics card may take several hundreds watts in average during model training, while an embedded GPU device may take only couple watts for the same purpose at the cost of a longer training time. In this paper, we report our empirical study on the model training using NVIDIA RTX 2080 Ti graphics card and NVIDIA Jetson Nano embedded device. We show that, surprisingly, while the training time using the Jetson Nano is 30 times slower than that using the graphics card, the total energy consumption by Jetson Nano is actually only half. The result suggests that when the response time is less critical, one may choose to do model training on GPU embedded devices instead.

KW - Deep Learning

KW - Embedded Systems

KW - Energy Efficiency

UR - https://www.scopus.com/pages/publications/85107232306

UR - https://www.scopus.com/pages/publications/85107232306#tab=citedBy

U2 - 10.1145/3450268.3453507

DO - 10.1145/3450268.3453507

M3 - Conference contribution

AN - SCOPUS:85107232306

T3 - IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation

SP - 253

EP - 254

BT - IoTDI 2021 - Proceedings of the 2021 International Conference on Internet-of-Things Design and Implementation

PB - Association for Computing Machinery, Inc

Y2 - 18 May 2021 through 21 May 2021

ER -

Energy Efficiency and Timeliness in Model Training for Internet-of-Things Applications: Poster Abstract

Abstract

Publication series

Conference

UN SDGs

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this