Network Management for Quality of Service Provisioning in Dynamic Wireless Networks

Abstract

As wireless network technologies develop, the performance of wireless networks is getting closer to wired networks and the number of users using wireless networks is increasing every year. In particular, as Internet of things (IoT) technology that connects all things through Internet becomes popular, numerous IoT devices, such as drones, vehicles, robots, and home appliances, are connected with each other over wireless networks to provide a variety of application services to people.

In this dissertation, we consider wireless networks through which drones are connected with remote computing servers or nearby drones to provide various application services, such as surveillance, object tracking, or wireless network services for disaster situations, to people. In the wireless networks that serve drones, states of wireless connections between drones and remote computing servers or among drones can be dynamically changed due to the mobility of drones or the wireless interference. These changes result in poor QoS in the wireless networks. To provide high-quality services to traffic, software-defined networking (SDN) technology can be used, which is a mechanism that separates a network into a data plane serving the user traffic and a control plane serving the control traffic. In SDN, since a centralized SDN controller is connected to each network device through a control plane, it may be possible to quickly detect changes in wireless networks and reconfigure the network to ensure the QoS of traffic. In this dissertation, we focus on network management technologies for quality of service (QoS) provisioning in SDN-based back-end radio access networks of drones and SDN-based wireless mesh networks among drones.

In the first study, we deal with a drone controller migration scheme that reduces network delays between drones and remote drone flight controllers in SDN-based radio access networks. We consider an environment where drone controllers are executed in wireless base stations (BSs) with cloud computing capabilities and drones are remotely connected to the drone controllers via wireless communication. As a drone moves away from a BS where its controller is executed, network delays between the drone and its controller can largely increase, which results in the unstable control of the drone. We propose a seamless drone controller migration scheme to migrate drone controllers to BSs close to drones for real-time drone control.

In the second study, we deal with an interference management scheme that reduces the drop rate of control planes in single-channel-based SDN drone mesh networks. In the single-channel-based SDN, the control and data plane traffic is transmitted through the same wireless channel. As the amount of data-plane traffic increases, the wireless interference the data plane imposes on the control plane increases and control packets can be dropped. If lots of important control packets are dropped, the control-plane connections between the a drone where an SDN controller is executed and its nearby drones becomes unstable and the drone mesh network can be out-of-control. We propose a control-plane interference management scheme that guarantees the QoS of the control plane by adjusting the amount of interference the data plane imposes on the control plane to operate the drone mesh network stably.

In the third study, we deal with an interference management scheme that differentiates the QoS of data-plane traffic according to priorities in multi-channel-based SDN drone mesh network. In the multi-channel-based SDN, since dedicated wireless channels can be allocated to a control plane, the QoS of the control plane is not affected by the wireless interference from the data plane. However, there is still the wireless interference within the data plane, which degrades the QoS of data-plane traffic significantly. Data-plane traffic processed in a drone mesh network may have various QoS requirements depending on the characteristic of traffic or the type of application. We propose a prioritized interference management scheme that differentiates the QoS of data plane traffic according to priorities by allowing each traffic to be serviced in different virtual networks depending on their QoS requirements and routing flows belonging to each slice differently.