Spectral Clustering-Based Controller Placement Approach in Low-Earth-Orbit Satellite Networks

Abstract

Because of their low latency, cost-effective construction, and rapid product updating, low-Earth-orbit (LEO) satellite networks have become an important part of the space infrastructure system. Software-defined network technology provides a candidate solution to adapt quickly and flexibly to network conditions and users’ requirements. However, with the numerical increase in satellites and expansion on the network scale, the multiple controller placement and the division of network control rights (MCP-DCR) problem in software-defined-network-enabled LEO satellite networks is a tough challenge with limited onboard resources and high dynamic topology. Aiming at the MCP-DCR problem in LEO satellite networks, this paper proposes a spectral clustering-based controller placement approach (SCCP) with reliability, latency, and load considered. In the proposed approach, a similarity matrix is adopted to cluster satellites firstly, and then a reliability model is constructed in which partial switch nodes are enabled with programmability. Finally, the network utility function is used to comprehensively measure network performance, and multiple controllers are determined based on network utility function optimization. Simulation results validate the effectiveness and adaptation of the proposed SCCP in multiple LEO satellite networks, which can improve the reliability and latency of controllers compared to those of the K-means-based method and exhaustive method.