Pod-Orchestrated Dual Cluster Framework for Intelligent Task Offloading in Satellite-Ground Edge Networks

This paper proposes a Kubernetes-based dualcluster framework for intelligent task offloading in satelliteground edge networks. The system integrates 12 ground stations and 66 satellite nodes, leveraging containerized pods and realtime monitoring to optimize task placement based on CPU usage, network load, and line-of-sight availability. A custom Python controller dynamically enforces offloading and migration policies, while Prometheus and Grafana facilitate continuous performance tracking. Simulation results show that the framework reduces end-to-end latency by up to $35 \%$, decreases CPU utilization on overloaded ground nodes by $43 \%$, and improves task success rates to over $90 \%$ within the optimal offloading range. These outcomes validate the framework’s scalability and resilience in managing dynamic workloads across heterogeneous terrestrial and non-terrestrial infrastructures.