Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices

Ziyang Zhang; Yang Zhao; Jie Liu

doi:10.1145/3583120.3589819

Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices

Ziyang Zhang
, Yang Zhao
, Jie Liu

Harbin Institute of Technology Shenzhen

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

3 Scopus citations

Abstract

Due to resource constraints, it is challenging to optimize the inference performance in terms of energy consumption and latency on edge devices. In this paper, we leverage both the dynamic voltage frequency scaling (DVFS) technique and edge-cloud collaborative inference to minimize the overall energy consumption. We propose a deep reinforcement learning (DRL)-based method called DVFO to jointly optimize 1) CPU, GPU and memory frequencies, and 2) the ratio of offloaded feature maps in edge-cloud collaboration. Preliminary experimental results show that DVFO reduces the average energy consumption by 33% compared to the baselines. Moreover, it reduces the inference latency by more than 54%.

Original language	English
Title of host publication	IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks
Publisher	Association for Computing Machinery, Inc
Pages	304-305
Number of pages	2
ISBN (Electronic)	9798400701184
DOIs	https://doi.org/10.1145/3583120.3589819
State	Published - 9 May 2023
Externally published	Yes
Event	22nd ACM/IEEE International Conference on Information Processing in Sensor Networks, IPSN 2023 - San Antonio, United States Duration: 9 May 2023 → 12 May 2023

Publication series

Name	IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks

Conference

Conference	22nd ACM/IEEE International Conference on Information Processing in Sensor Networks, IPSN 2023
Country/Territory	United States
City	San Antonio
Period	9/05/23 → 12/05/23

UN SDGs

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

SDG 7 Affordable and Clean Energy

Keywords

Collaborative Inference
DVFS
Edge Computing
Reinforcement Learning

Access to Document

10.1145/3583120.3589819

Cite this

Zhang, Z., Zhao, Y., & Liu, J. (2023). Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices. In IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks (pp. 304-305). (IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks). Association for Computing Machinery, Inc. https://doi.org/10.1145/3583120.3589819

Zhang, Ziyang ; Zhao, Yang ; Liu, Jie. / Poster Abstract : DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices. IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks. Association for Computing Machinery, Inc, 2023. pp. 304-305 (IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks).

@inproceedings{0280398f41bc4ed5bd0e1e4d4c521b49,

title = "Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices",

abstract = "Due to resource constraints, it is challenging to optimize the inference performance in terms of energy consumption and latency on edge devices. In this paper, we leverage both the dynamic voltage frequency scaling (DVFS) technique and edge-cloud collaborative inference to minimize the overall energy consumption. We propose a deep reinforcement learning (DRL)-based method called DVFO to jointly optimize 1) CPU, GPU and memory frequencies, and 2) the ratio of offloaded feature maps in edge-cloud collaboration. Preliminary experimental results show that DVFO reduces the average energy consumption by 33\% compared to the baselines. Moreover, it reduces the inference latency by more than 54\%.",

keywords = "Collaborative Inference, DVFS, Edge Computing, Reinforcement Learning",

author = "Ziyang Zhang and Yang Zhao and Jie Liu",

note = "Publisher Copyright: {\textcopyright} 2023 Owner/Author.; 22nd ACM/IEEE International Conference on Information Processing in Sensor Networks, IPSN 2023 ; Conference date: 09-05-2023 Through 12-05-2023",

year = "2023",

month = may,

day = "9",

doi = "10.1145/3583120.3589819",

language = "英语",

series = "IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks",

publisher = "Association for Computing Machinery, Inc",

pages = "304--305",

booktitle = "IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks",

}

Zhang, Z, Zhao, Y & Liu, J 2023, Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices. in IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks. IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks, Association for Computing Machinery, Inc, pp. 304-305, 22nd ACM/IEEE International Conference on Information Processing in Sensor Networks, IPSN 2023, San Antonio, United States, 9/05/23. https://doi.org/10.1145/3583120.3589819

Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices. / Zhang, Ziyang; Zhao, Yang ; Liu, Jie.
IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks. Association for Computing Machinery, Inc, 2023. p. 304-305 (IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks).

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

TY - GEN

T1 - Poster Abstract

T2 - 22nd ACM/IEEE International Conference on Information Processing in Sensor Networks, IPSN 2023

AU - Zhang, Ziyang

AU - Zhao, Yang

AU - Liu, Jie

PY - 2023/5/9

Y1 - 2023/5/9

N2 - Due to resource constraints, it is challenging to optimize the inference performance in terms of energy consumption and latency on edge devices. In this paper, we leverage both the dynamic voltage frequency scaling (DVFS) technique and edge-cloud collaborative inference to minimize the overall energy consumption. We propose a deep reinforcement learning (DRL)-based method called DVFO to jointly optimize 1) CPU, GPU and memory frequencies, and 2) the ratio of offloaded feature maps in edge-cloud collaboration. Preliminary experimental results show that DVFO reduces the average energy consumption by 33% compared to the baselines. Moreover, it reduces the inference latency by more than 54%.

AB - Due to resource constraints, it is challenging to optimize the inference performance in terms of energy consumption and latency on edge devices. In this paper, we leverage both the dynamic voltage frequency scaling (DVFS) technique and edge-cloud collaborative inference to minimize the overall energy consumption. We propose a deep reinforcement learning (DRL)-based method called DVFO to jointly optimize 1) CPU, GPU and memory frequencies, and 2) the ratio of offloaded feature maps in edge-cloud collaboration. Preliminary experimental results show that DVFO reduces the average energy consumption by 33% compared to the baselines. Moreover, it reduces the inference latency by more than 54%.

KW - Collaborative Inference

KW - DVFS

KW - Edge Computing

KW - Reinforcement Learning

UR - https://www.scopus.com/pages/publications/85160070118

U2 - 10.1145/3583120.3589819

DO - 10.1145/3583120.3589819

M3 - 会议稿件

AN - SCOPUS:85160070118

T3 - IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks

SP - 304

EP - 305

BT - IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks

PB - Association for Computing Machinery, Inc

Y2 - 9 May 2023 through 12 May 2023

ER -

Zhang Z, Zhao Y , Liu J. Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices. In IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks. Association for Computing Machinery, Inc. 2023. p. 304-305. (IPSN 2023 - Proceedings of the 2023 22nd International Conference on Information Processing in Sensor Networks). doi: 10.1145/3583120.3589819

Poster Abstract: DVFO: Dynamic Voltage, Frequency and Offloading for Efficient AI on Edge Devices

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this