Dynamic SDN Controller Placement in Multi-Layer Satellite Network

The evolution of global satellite networks into multi-layered and large-scale systems has enabled significant improvements in coverage reliability, capacity scalability, and service flexibility. Nevertheless, the coexistence of highly dynamic network topology and spatiotemporal traffic heterogeneity requires adaptive controller placement to address the challenges of effective network management and control. In this paper, we study the dynamic software-defined networking (SDN) controller placement problem in multi-layer satellite networks and introduce a two-stage optimization framework, aiming to determine optimal controller number, locations, and switch-controller mappings under dynamic topology and traffic variations. Firstly, a dynamic network partitioning method based on the Louvain algorithm is proposed to optimize both quality of service (QoS) and load balancing under constraints of limited bandwidth. Then, a hybrid genetic algorithm with greedy initialization is employed to minimize intra-domain and inter-domain delays through efficient controller placement. Additionally, a mathematical model for the data plane is introduced, incorporating a realistic traffic profile that captures spatial and temporal variations in user demands. Simulation results demonstrate that the proposed approach outperforms benchmarks in terms of delay and load balancing.