Research On Availability Of 5G Network Slices

As the core technology of 5G,network slicing solves the challenges of the access of massive mobile devices and the emergence of various application scenarios.Availability as an important indicator to measure the quality of network services.Too low availability can lead to a serious decline in service quality.This thesis focuses on the problems of the availability of 5G network slices.We choose the promising Wireless-Optical Broadband Access Network(WOBAN)as the network architecture for this research,and focusing on"the problem of high-availability slice capacity" and "the problem of provisioning availability-guaranteed slices".Firstly,we study the problem of high-availability slice capacity based on WOBAN.For this problem,we consider two network slice provisioning scenarios,i.e.,fully provisioned slices and partially provisioned slices.The fully provisioned slice means that the resource requirement of all virtual links and virtual nodes of the slice is satisfied,while the partially provisioned slice means that the resource requirement of some virtual links and virtual nodes of the slice is satisfied.To maximize the availability-weighted capacity of network slices,we develop Integer Linear Programming(ILP)models and propose efficient heuristic algorithms.The simulation results show that the partially provisioned slice can provide more availability-weight capacity than the fully provisioned slice.Secondly,we study the problem of provisioning availability-guaranteed slices based on WOBAN.For this problem,we propose a new availability definition for the network slice.Specifically,we consider four types of network fault scenarios,including fiber link failure,microwave link failure,physical node failure,and Optical Line Terminal(OLT)failure.In order to maximize the number of availability-guaranteed slices,we develop an ILP model and also propose an efficient heuristic algorithm that can balances the network traffic load and maximizes the slice availability.The simulation results show the efficiency of the proposed methods.