Fault-Tolerant Scheduling Mechanism for Dynamic Edge Computing Scenarios Based on Graph Reinforcement Learning

Area of Science:

• Computer Science
• Artificial Intelligence
• Distributed Systems

Background:

• Edge computing handles real-time processing for Internet of Things (IoT) devices.
• Edge environments face challenges like network instability and dynamic resource variations.
• These challenges can cause task interruptions and system failures.

Purpose of the Study:

• To develop a fault-tolerant scheduling mechanism for edge computing environments.
• To ensure continuous system operation despite node failures.
• To enhance the robustness of edge computing systems.

Main Methods:

• Proposed an innovative fault-tolerant scheduling model based on asynchronous graph reinforcement learning.
• Utilized a deep reinforcement learning framework with a graph neural network to model node communication.
• Implemented an asynchronous model update strategy for real-time dynamic scheduling.

Main Results:

• The proposed model accurately captured complex communication relationships between computing nodes.
• The model generated effective fault-tolerant scheduling actions for dynamic environments.
• Experimental results showed superior performance compared to baseline algorithms.

Conclusions:

• The asynchronous graph reinforcement learning model enhances fault-tolerant scheduling in edge computing.
• The method significantly improved quality of service (QoS) assurance.
• The approach offers a robust solution for dynamic and unstable edge environments.