Unified Optimization Framework for Distributed LEO-assisted SD-ISTN: Resilient and Scalable Solution for Dynamic Controller Placement

Abstract

Low Earth Orbit (LEO)-assisted Software-Defined Integrated Satellite-Terrestrial Networks (SD-ISTNs) offer worldwide, adaptable connectivity. However, scalability constraints, dynamic network topologies, and inconsistent traffic loads in LEO constellations highlight the importance of addressing the controller placement problem (CPP), which is a critical design problem that influences signaling plane latency and overall utilization of resources. This paper presents a resilient and scalable solution to the CPP in SD-ISTNs by formulating it as a multi-metric optimization problem aimed at concurrently minimizing key latency components, queuing delay, and controller processing overhead. A two-stage approach is proposed to achieve efficient resolution. In the first stage, a clustering-based cell association technique is employed for grouping satellite switches and user equipment (UE) into delay-constrained sub-clusters, informed by the network load demands and latency constraints. This process generates load and latency profiles for each LEO satellite, which are critical inputs for the second stage. In the second stage, a weighted-sum scalarization is utilized to convert the multi-objective problem into a single-objective formulation while preserving Pareto efficiency. This facilitates trade-offs among competing parameters, including latency components, queuing delay, and controller processing overhead. Through comprehensive simulations, we demonstrate that the proposed approach surpasses state-of-the-art methods, achieving up to 48% lower flow setup time, 50% lower packet drop rate, and maintaining fairness indices above 0.85 under dynamic traffic conditions, thereby offering a resilient and scalable solution for CPP in dynamic LEO-assisted SD-ISTNs.