Dynamic Resource Coordination towards Reliable and Flexible Network Slicing




Journal Title

Journal ISSN

Volume Title




The increasing demand for a diverse array of network applications entails a more flexible and reliable networking paradigm. Network slicing is envisioned to package a set of networking, computing, and storage resources in a coordinated manner, so that the network slice with the tailored set of resources satisfies service requirements unique to each network application. A key enabling technology of network slicing is network softwarization, exemplified by Software Defined Networking (SDN) and Network Function Virtualization (NFV), which enable swift migration of both networking and computing resources. While network slicing sets a conceptual foundation for next-generation networking for diverse applications, the resource coordination mechanism that dynamically operates and manages the resources remains a challenging issue. This is more so when considering the computational complexity induced by the dependency relationship among the softwarized resources and the uncertainty of future network states, such as network failure scenarios and traffic patterns. This dissertation features four resource allocation problems that collectively facilitate the agile operation and management of end-to-end network slices. The first two problems are related to the reliability aspect of network slicing. In particular, we discuss protection and recovery problems of interdependent network components from a resource allocation standpoint. The third problem deals with a dynamic bandwidth allocation in optical access networks. The dynamic adjustment of allocated bandwidth assists more effective resource utilization and the accommodation of different types of network services. Furthermore, SDN controller placement, which determines responsiveness to a request for end-to-end resource coordination, is examined as the fourth problem. The theoretical analysis and proposed algorithms for the problems not only solve the specific resource allocation tasks, but also provide fundamental insights to tackle similar allocation problems under entangled dependency and future uncertainty. In particular, the proposed learning-based approaches project an automated resource coordination system for more effective utilization of network resources in 5G and beyond networks.



Resource allocation, Computer networks, End-to-end delay (Computer networks), Resource allocation