DNN-driven task scheduling for high performance edge-cloud heterogeneous systems
Uddin, Md Raihan Asaduzzaman, Abu Nawar, Fairuz Thompson, Christian C.
Uddin, Md Raihan
Asaduzzaman, Abu
Nawar, Fairuz
Thompson, Christian C.
An error occurred retrieving the object's statistics
Citations
Altmetric:
Authors
Uddin, Md Raihan
Asaduzzaman, Abu
Nawar, Fairuz
Thompson, Christian C.
Asaduzzaman, Abu
Nawar, Fairuz
Thompson, Christian C.
Other Names
Location
Time Period
Advisors
Original Date
Digitization Date
Issue Date
2025-10-16
Type
Conference paper
Genre
Keywords
Edge-cloud computing,Energy consumption,Execution time,Heterogeneous systems,ML models
Subjects (LCSH)
Citation
Uddin, M. R., Asaduzzaman, A., Nawar, F., & Thompson, C. (2025). DNN-driven task scheduling for high performance edge-cloud heterogeneous systems. 15-19 Sept. 2025.
Abstract
Integrating Artificial Intelligence (AI) into Internet of Things (IoT) devices has significantly increased computational demands and reduced latency and energy efficiency requirements on Cloud Servers (CSs). Edge Servers (ESs) address latency and energy efficiency requirements, but introduce the complexity of optimal task scheduling among servers. Existing solutions often prioritize energy or latency reduction while overlooking critical factors such as server utilization to optimize task scheduling. In this work, we propose a deep learning-driven task scheduling framework for heterogeneous systems based on the edge cloud to improve performance. A heterogeneous system comprising 40 IoT devices, one CS, and four ESs is evaluated with varying task sizes from 2 GiB to 10GiB. Each ES is employed with a Deep Neural Network (DNN) for optimizing task scheduling to minimize execution time, energy consumption, and CS utilization simultaneously. Comparative analysis demonstrates that our approach achieves up to 21.9% and 275.4% faster execution times compared to the Optimal Pairing Ratio (OPR) method and traditional cloud-only methods, respectively. Energy consumption is reduced by up to 12.8% over OPR and 112.1% over the cloud-only method, respectively. Additionally, CS utilization is reduced by 7.2% over OPR and 38.4% over the cloud-only method. This study can be further extended to improve the scalability of the edge-cloud heterogeneous system.
Table of Contents
Description
Click on the DOI link to access this conference paper at the publishers website (may not be free).
Publisher
Institute of Electrical and Electronics Engineers Inc.
Journal
Book Title
Series
2025 IEEE High Performance Extreme Computing Conference (HPEC)
Digital Collection
Finding Aid URL
Use and Reproduction
Archival Collection
PubMed ID
ISSN
2643-1971
2377-6943
2377-6943