UCLA

Cellular networks are the only large-scale system that supports mobility. It has successfully provided anytime and anywhere connectivity by deploying dense cell towers and switching the serving cell when the client leaves its coverage. To meet stringent network service requirements, 5G/4G has been actively enhancing network capabilities (e.g., mmWave bands, carrier aggregation, multi-carrier access). Mobility support also plays a critical role in achieving high reliability and throughput. They are desired not only in low mobility but also in extreme mobility; extreme mobility has become more common given high-speed rails and mmWave cells with a small coverage. However, the current practice fails to address the challenges. The culprit is that legacy design is confined to connectivity and can hardly match upgraded network capability and demand. This dissertation showcases reliability and throughput issues in extreme and low mobility. For reliability, failures increase vastly in extreme mobility. Multi-mobile-carrier access (e.g., Google Fi) incurs oscillation among carriers akin to BGP loops. Furthermore, mobility support largely misses cells with high throughput despite the excellent potential of carrier aggregation and rich radio resources. Throughput loss is more considerable with mmWave cells due to less decision-making time. This dissertation proposes revolutionary solutions regarding the limitation of legacy mobility support design: 1. We enhance network reliability. We devise REM, the first movement-based mobility management, to combat dramatic wireless dynamics and mitigate failures in extreme mobility. In addition, we eliminate inter-carrier switching loops by resolving conflicts in multiple parties' policies. 2. Our design boosts throughput via better utilization of increased radio resources and advanced technologies. We devise CA++ to fulfill CA potentials over a broad frequency spectrum, including mmWave cells. It reforms the conventional cell-by-cell selection into the parallel design, which relaxes the dilemma of making a rapid or good selection. We design RPerf, a ready-to-launch algorithm to enable throughput-aware cell selection with reconfiguration.