User-centric Throughput/Latency Enhancement in Wired/Wireless Networks

Abstract

Historically, there have been close interactions between communication technologies and network applications. With the evolution ofcommunications, user applications have been expanding from web browsing, social network service, and video streaming to the soon-to-come tactile Internet, real-time interactive Virtual/Augmented Reality (VR/AR), and remote medical care. In turn, such future applications require what the current networks cannot guarantee, leading to the need for next-generation networking technologies with improved performance. Since the emerging applications are expected to deliver multi-dimensional and multi-sensory information, Quality-of-Experience (QoE), which describes the end user’s overall satisfaction with the service experience, will become a key factor in evaluating how successful a network service is. In order to provide better QoE, next-generation networks should enhance the user-centric performance metrics like user experienced throughput, latency, and reliability, as well as the traditional metrics like system-level throughput. This dissertation introduces a series of approaches to improve user-centric throughput/latency in wired/wireless networks. First, to enhance per-user throughput of WLAN by mitigating the classical phenomenon called performance anomaly, we propose co-locating a Super Wi-Fi access point (AP) with a Wi-Fi AP and then relocating low-rate Wi-Fi stations (STAs) to Super Wi-Fi in theoutskirts of Wi-Fi’s coverage. Next, two flit scheduling algorithms for the cut-through switching wired network have been developedto achieve low end-to-end latency with high reliability. Finally, novel latency-aware MU-MIMO scheduling algorithms are introduced to enhance the user experienced latency in WLAN-based VR systems. Via extensive evaluations including numerical analysis and computer simulations, it is verified that the proposed schemes can enhance the user-centric performance.