(A) distributed system for the mobility of everything

Abstract

Everything can be mobile, from end-hosts to applications (or processes). Today, more and more end-hosts that include mobile devices such as smartphones are being connected with IEEE 802.11 wireless LANs (WLANs or Wi-Fi). However, mobile users are experiencing poor service quality due to the large handoff delay when they roam between different wireless Access Points (APs). The large handoff delay causes the delay in application services, especially in real-time multimedia services, such as Voice over IP (VoIP) and video conferencing. This type of mobility problem is generally categorized as link-layer (Layer 2) mobility. When roaming between subnets, end-hosts and applications are experiencing changes in their IP addresses. This results in additional handoff delay and even causes connectivity losses on the application services. In the literature, this type of mobility problem belongs to network-layer (Layer 3) mobility. In order to provide the mobility of possible things, the link and network-layer mobility is studied in this dissertation. First, as the link-layer mobility, a new handoff scheme called Directional Handoff (DH) is proposed, which uses the geomagnetic sensor embedded in mobile devices. The proposed DH scheme predicts the movement direction of a Mobile Station (MS) from the currently associated AP and performs active scanning with a reduced number of channels. Second, as the network-layer mobility, a new IP mobility approach called Mobility of Everything (MoE) is proposed. The need for IP mobility is arising more and more due to the increasing number of mobile devices that embed two or more wireless connectivity technologies. Meanwhile, the Virtual Machine (VM) technology is being applied to native applications, which enables the live migration of an application as an alternative to the VM mobility. The proposed MoE approach unifies possible mobility cases by using a distributed mobility system. The approach uses a hash value as an object ID, which is used to identify the specific edge switch where the object is connected. Furthermore, the approach distributes the binding information of an object ID and IP addresses throughout the network. Over the proposed mobility approaches, a retransmission scheme called Lightweight Retransmission Protocol (LRP) is proposed. The packet loss problem that occurs during the link and network-layer handoff procedures is another cause of restricting user mobility because the packet losses degrade the quality of video. Therefore, in order to improve the quality of real-time multimedia during the handoff procedures, the proposed LRP is needed in conjunction with the link and network-layer handoff procedures. The proposed DH scheme is implemented on commercial Android smartphones, and its performance is evaluated in a real indoor WLAN environment. The experimental results demonstrate that the DH scheme reduces the handoff delay compared to the conventional handoff and selective scanning scheme. In addition, the proposed MoE approach is implemented on OpenFlow-enabled software switches, and its feasibility is evaluated in an emulation environment. The evaluation results demonstrate that the MoE approach achieves seamless handoff of end-hosts and applications while maintaining their IP addresses. It also achieves the scalability in terms of the number of concurrent mobile objects. Finally, the proposed LRP in conjunction with the DH scheme maintains quality for real-time video even in an environment with frequent link-layer handoffs.