Study On Survivable Network Slicing Algorithms In 5G Core Networks

Network slicing is a key technology in 5G networks.Its essence is to divide a shared infrastructure network into multiple logically independent virtual networks to provide on-demand support for different services.Network slicing enables the flexible allocation of network resources and improve the resource utilization ratio.However,in real networks,node or link failures occur occasionally,and it is impossible to avoid them absolutely.When one node or link failure occurs,those virtual networks onto it will be affected,which results in the degradation of Quality of Service(Qo S),and even the service interruption.This is unacceptable for both operators and users.In order to cope with occasional network failures,it is usually necessary to reserve a large quantity of backup resources for restoring those virtual networks affected.However,it will make a great number of idle resources in networks unable to be effectively used,which leads to a significant reduction in the resource utilization ratio and increases the cost of operators.Therefore,it is important for us to ensure the survivability of network slices and minimize the required backup resources.This thesis studies survivable network slicing algorithms in 5G core networks for single link failure and node failure respectively.Our main work is as follows:1.For single link failure,we propose a multi-path shared backup(MPSB)algorithm.This algorithm improves the resource utilization ratio through a sharing mechanism and takes Qo S into account.Firstly,a existing multi-path mapping scheme is improved in this thesis.Each virtual link is mapped into multiple disjoint physical paths,and primary and backup bandwidth are allocated on each path to reduce the backup bandwidth needed in our scheme.Secondly,the sufficient condition for sharing the backup bandwidth among the mapped path sets of different virtual links is obtained through our research,so that the required backup resources can be further reduced while a strong survivability is provided.Considering that multi-path mapping will generate additional network overhead,and result in the degradation of Qo S,the penalty cost based on the maximum path delay difference of the mapped path set is formulated.Then,the penalty cost is added into the optimization model of survivable virtual link mapping to seek a great balance between the resource utilization ratio and Qo S.In order to shorten the time of calculation,a multi-path shared backup(MPSB)algorithm is proposed to acquire the sub-optimal solution.Finally,we evaluate the performance of the MPSB algorithm in different simulation scenarios.Simulation results show that the MPSB algorithm reduce the backup bandwidth ratio to20%-30%,and its performance is better than existing algorithms.Furthermore,the MPSB algorithm achieve the conversion between the resource utilization and Qo S by adjusting the weight parameters of the penalty cost.2.For single node failure,this thesis puts forward a shared backup network slicing(SBNS)algorithm,which jointly implements node and link mapping,as well as primary and backup mapping,and reduces these required resources through sharing backup resources.Firstly,a shared backup scheme applying to network slicing is proposed by referring to the existing sharing mechanisms.Multiple virtual network functions(VNFs)share the same backup VNF,and the primary and backup path of each virtual link share the same bandwidth to reduce backup resources needed in our scheme.Based on this scheme,the network slicing problem is modeled as a mixed integer linear programming(MILP).It can greatly reduce the required resources while strong survivability is provided for network slices.Finally,we evaluate the performance of the SBNS algorithm in different simulation scenarios.The simulation results show that the SBNS algorithm reduces the backup resource ratio to 17%-25%,which significantly improves the resource utilization ratio.