Transport protocol design for end-to-end data delivery in emerging wireless networks
Fan Zhou · 2019
Driven by the rapidly increasing popularity of Internet-connected phones, tablets and wearable devices, new wireless technologies and standards are emerging at an unprecedented pace. The evolution of wireless networking opens a myriad of possible mobile applications, which have greatly transformed the way we access information, entertain and connect with others. A critical question is how to satisfy the quality of service (QoS) requirements for various applications by delivering the data to users reliably and efficiently under different networking conditions. While many protocols have been proposed for better QoS, they usually requires modification to the networking core or heavy interaction between the layers of the networking stack. Therefore, the goal of dissertation is to develop lightweight, scalable and end-to-end transport solutions to bridge the long-standing gap between the increasing diversity of the mobile applications and the constant evolving wireless technologies. We first focus on studying how to tweak the logic and functionality of traditional TCP protocol given a new wireless networks with different link properties in the context of cognitive radio (CR) network. We start with an extensive ns-3 simulation study of three different classical end-to-end TCP flavors. We conclude that the main adverse impact to the TCP connection in a CR network is the interruption during the transmission period caused by spectrum sensing or primary user arrival events. Based on this insight, we propose TCP C^2, a method that greatly improves the flow responsiveness to abrupt variation of underlying layer spectrum availability in cellular CR architectures. We show that C^2 achieves 2x better throughput compared to TCP NewReno and TCP Westwood+, and approximately 20% improvement in throughput while also incurring 4X shorter queuing delay against Cubic (default implementation in Linux). Our second contribution is Janus, a new transport-layer framework that automatically selects among existing congestion control variants to optimize traffic in accordance with application demands. Janus solves the problem of how to select the right protocol to optimize the QoS for different applications given a large collection of existing TCP variants. We implement Janus in the Linux kernel and extensively evaluate its performance with both emulated and real Internet traffic. Our results show that Janus can significantly improve user-perceived performance according to QoE metrics, with up to 5X fewer interruptions for video streaming applications and 2X faster page loading for web-browsing applications. Finally, we seek to answer the question of how to increase the robustness of data delivery in mmWave networks with densely deployed access points (AP). We conduct real-world experiments with off- the-shelf 60GHz devices and show that naively connecting to any available AP, or even multiple APs, may not fully realize the promise of efficiently utilizing the extremely large bandwidths available in this band. To tackle this problem, we propose Multi-AP Association Protocol (MAP). MAP is a cross-layer connection management mechanism that runs only at the client-side and does not need centralized support from the APs, thus avoiding overheads of added control signaling. We show that MAP can substantially increase the QoS for videostreaming applications for mmWave networks with testbed evaluation and large-scale simulation.