| Description |
The next generation mobile network (i.e., 5G network) is expected to host emerging use cases that have a wide range of requirements; from Internet of Things (IoT) devices that prefer low-overhead and scalable network to remote machine operation or remote healthcare services that require reliable end-to-end communications. Improving scalability and reliability is among the most important challenges of designing the next generation mobile architecture. The current (4G) mobile core network heavily relies on hardware-based proprietary components. The core networks are expensive and therefore are available in limited locations in the country. This leads to a high end-to-end latency due to the long latency between base stations and the mobile core, and limitations in having innovations and an evolvable network. Moreover, at the protocol level the current mobile network architecture was designed for a limited number of smart-phones streaming a large amount of high quality traffic but not a massive number of low-capability devices sending small and sporadic traffic. This results in high-overhead control and data planes in the mobile core network that are not suitable for a massive number of future Internet-of-Things (IoT) devices. In terms of reliability, network operators already deployed multiple monitoring sys- tems to detect service disruptions and fix problems when they occur. However, detecting all service disruptions is challenging. First, there is a complex relationship between the network status and user-perceived service experience. Second, service disruptions could happen because of reasons that are beyond the network itself. With technology advancements in Software-defined Network (SDN) and Network Func- tion Virtualization (NFV), the next generation mobile network is expected to be NFV-based and deployed on NFV platforms. However, in contrast to telecom-grade hardware with built-in redundancy, commodity off-the-shell (COTS) hardware in NFV platforms often can't be comparable in term of reliability. Availability of Telecom-grade mobile core network hardwares is typically 99.999% (i.e., "five-9s" availability) while most NFV platforms only guarantee "three-9s" availability - orders of magnitude less reliable. Therefore, an NFV-based mobile core network needs extra mechanisms to guarantee its availability. This Ph.D. dissertation focuses on using SDN/NFV, data analytics and distributed system techniques to enhance scalability and reliability of the next generation mobile core network. The dissertation makes the following contributions. First, it presents SMORE, a practical offloading architecture that reduces end-to-end latency and enables new functionalities in mobile networks. It then presents SIMECA, a light-weight and scalable mobile core network designed for a massive number of future IoT devices. Second, it presents ABSENCE, a passive service monitoring system using customer usage and data analytics to detect silent failures in an operational mobile network. Lastly, it presents ECHO, a distributed mobile core network architecture to improve availability of NFV-based mobile core network in public clouds. |
