CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System

Zimo Ma; Yifeng Li; Di Liu; Ziheng Zhang; Kuangyu Zheng

doi:10.1145/3587716.3587744

CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System

Zimo Ma
, Yifeng Li
, Di Liu
, Ziheng Zhang
, Kuangyu Zheng^*

^*Corresponding author for this work

School of Electronic and Information Engineering

Beihang University

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

Abstract

In this paper, we investigate both the energy consumption and running time of different CNN tasks on GPUs or CPUs, and analyze their characterization for different CNN models under different application and system configuration factors. We find that this joint energy consumption and makespan optimization problem can be formulated as an integer linear programming problem. Then we propose CHESS (CNN-task Heterogeneous Efficient Scheduling System) with a two-stage heuristic scheduling algorithm, to better allocate computing resources for the upcoming tasks, and to schedule them dynamically on the heterogeneous cluster. Experiments show that our CHESS can save up to 15.9% energy and decrease up to 32.7% makespan over existing approaches.

Original language	English
Title of host publication	ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing
Publisher	Association for Computing Machinery
Pages	172-176
Number of pages	5
ISBN (Electronic)	9781450398411
DOIs	https://doi.org/10.1145/3587716.3587744
State	Published - 17 Feb 2023
Event	15th International Conference on Machine Learning and Computing, ICMLC 2023 - Hybrid, Zhuhai, China Duration: 17 Feb 2023 → 20 Feb 2023

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	15th International Conference on Machine Learning and Computing, ICMLC 2023
Country/Territory	China
City	Hybrid, Zhuhai
Period	17/02/23 → 20/02/23

UN SDGs

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

SDG 7 Affordable and Clean Energy

Keywords

CNN
characteristic analysis
dynamic task scheduling
efficient computing
heterogeneous system

Access to Document

10.1145/3587716.3587744

Cite this

Ma, Z., Li, Y., Liu, D., Zhang, Z., & Zheng, K. (2023). CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System. In ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing (pp. 172-176). (ACM International Conference Proceeding Series). Association for Computing Machinery . https://doi.org/10.1145/3587716.3587744

@inproceedings{3e0e9c57336e4cd4bf784a186e14fe81,

title = "CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System",

abstract = "In this paper, we investigate both the energy consumption and running time of different CNN tasks on GPUs or CPUs, and analyze their characterization for different CNN models under different application and system configuration factors. We find that this joint energy consumption and makespan optimization problem can be formulated as an integer linear programming problem. Then we propose CHESS (CNN-task Heterogeneous Efficient Scheduling System) with a two-stage heuristic scheduling algorithm, to better allocate computing resources for the upcoming tasks, and to schedule them dynamically on the heterogeneous cluster. Experiments show that our CHESS can save up to 15.9\% energy and decrease up to 32.7\% makespan over existing approaches.",

keywords = "CNN, characteristic analysis, dynamic task scheduling, efficient computing, heterogeneous system",

author = "Zimo Ma and Yifeng Li and Di Liu and Ziheng Zhang and Kuangyu Zheng",

note = "Publisher Copyright: {\textcopyright} 2023 ACM.; 15th International Conference on Machine Learning and Computing, ICMLC 2023 ; Conference date: 17-02-2023 Through 20-02-2023",

year = "2023",

month = feb,

day = "17",

doi = "10.1145/3587716.3587744",

language = "英语",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery ",

pages = "172--176",

booktitle = "ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing",

address = "美国",

}

Ma, Z, Li, Y, Liu, D, Zhang, Z & Zheng, K 2023, CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System. in ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing. ACM International Conference Proceeding Series, Association for Computing Machinery , pp. 172-176, 15th International Conference on Machine Learning and Computing, ICMLC 2023, Hybrid, Zhuhai, China, 17/02/23. https://doi.org/10.1145/3587716.3587744

CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System. / Ma, Zimo; Li, Yifeng; Liu, Di et al.
ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing. Association for Computing Machinery , 2023. p. 172-176 (ACM International Conference Proceeding Series).

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

TY - GEN

T1 - CHESS

T2 - 15th International Conference on Machine Learning and Computing, ICMLC 2023

AU - Ma, Zimo

AU - Li, Yifeng

AU - Liu, Di

AU - Zhang, Ziheng

AU - Zheng, Kuangyu

PY - 2023/2/17

Y1 - 2023/2/17

N2 - In this paper, we investigate both the energy consumption and running time of different CNN tasks on GPUs or CPUs, and analyze their characterization for different CNN models under different application and system configuration factors. We find that this joint energy consumption and makespan optimization problem can be formulated as an integer linear programming problem. Then we propose CHESS (CNN-task Heterogeneous Efficient Scheduling System) with a two-stage heuristic scheduling algorithm, to better allocate computing resources for the upcoming tasks, and to schedule them dynamically on the heterogeneous cluster. Experiments show that our CHESS can save up to 15.9% energy and decrease up to 32.7% makespan over existing approaches.

AB - In this paper, we investigate both the energy consumption and running time of different CNN tasks on GPUs or CPUs, and analyze their characterization for different CNN models under different application and system configuration factors. We find that this joint energy consumption and makespan optimization problem can be formulated as an integer linear programming problem. Then we propose CHESS (CNN-task Heterogeneous Efficient Scheduling System) with a two-stage heuristic scheduling algorithm, to better allocate computing resources for the upcoming tasks, and to schedule them dynamically on the heterogeneous cluster. Experiments show that our CHESS can save up to 15.9% energy and decrease up to 32.7% makespan over existing approaches.

KW - CNN

KW - characteristic analysis

KW - dynamic task scheduling

KW - efficient computing

KW - heterogeneous system

UR - https://www.scopus.com/pages/publications/85173860840

U2 - 10.1145/3587716.3587744

DO - 10.1145/3587716.3587744

M3 - 会议稿件

AN - SCOPUS:85173860840

T3 - ACM International Conference Proceeding Series

SP - 172

EP - 176

BT - ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing

PB - Association for Computing Machinery

Y2 - 17 February 2023 through 20 February 2023

ER -

Ma Z, Li Y, Liu D, Zhang Z, Zheng K. CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System. In ICMLC 2023 - Proceedings of the 2023 15th International Conference on Machine Learning and Computing. Association for Computing Machinery . 2023. p. 172-176. (ACM International Conference Proceeding Series). doi: 10.1145/3587716.3587744

CHESS: Joint Energy and Makespan Optimization for Dynamic CNN Task Scheduling on the Heterogeneous System

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this