A GNN-DRL-based Collaborative Edge Computing Strategy for Partial Offloading

Edge computing is an emerging distributed computing paradigm that reduces computation latency and energy consumption by offloading application tasks from user devices to near-edge servers for execution. In order to utilize the computing resources of edge servers and increase efficiency, collaborativ...

Full description

Saved in:
Bibliographic Details
Main Authors: Tao, Tianhao, Hou, Jiacheng, Nayak, Amiya
Format: Conference Proceeding
Language:English
Subjects:
Collaboration
collaborative edge computing
Computational modeling
deep reinforcement learning
Edge computing
graph neural network
Network topology
partial computation offloading
Servers
Task analysis
Vectors
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Edge computing is an emerging distributed computing paradigm that reduces computation latency and energy consumption by offloading application tasks from user devices to near-edge servers for execution. In order to utilize the computing resources of edge servers and increase efficiency, collaborative edge computing is proposed as a new type of edge computing method where multiple edge servers can work together to solve a task. By dividing a task into interrelated subtasks, each subtask can be processed locally at the device or offloaded to an edge server to minimize processing latency. In the paper, we propose a GNN-DRL-based offloading strategy that considers the edge servers' task attributes and network topology to make an optimal offloading decision for each application subtask. Experiments show that our proposed method performs better than state-of-the-art and baseline strategies in reducing the average latency for each task.
ISSN:2576-6813
DOI:10.1109/GLOBECOM54140.2023.10436825