Research On The Resource Allocation Of Network Virtualization For 5G

Mobile communication will step into the age of fifth generation mobile communication system(5G),the three typical application scenarios of 5G need higher data transmission rate,more device connection density and lower end-to-end service delay and other key performance indicators.The resource allocation methods of network virtualization for 5G still have the following problems: 1)It is difficult to meet the requirement of 5G low-latency business,because lack of the service function chaining mapping method in one phase which can deal with the large number of services that are arriving dynamic.2)The redundancy backup method is used to ensure the reliability of the service,which does not fully consider the backup resource utilization and leads to the waste of backup resources.3)The virtual network function scheduling does not fully consider the link bandwidth allocation strategy,so that the transmission delay becomes the bottleneck which restricts the total service delay.So 5G introduced software defined network and network function virtualization.By decoupling control plane and user plane,network element function and special hardware,5G will realize the virtualization of infrastructure platform and can provide users with diversified network services on demand.In order to meet the requirements of 5G on performance indicators such as latency and reliability,efficient allocation of limited infrastructure resources is required.To cope with the above problems,this paper studied the resource allocation problem of network virtualization for 5G from 863 projects.This research first considered processing delay and transmission delay together,and found the service function chaining mapping scheme with the minimum delay cost of single network service by using dynamic programming;Then,by ordering the Backup-Cost Importance of the virtual network functions in the initial mapping view,the reliability of network services are improved using joint backup method;Finally,by changing the instantiation order of virtual network function in same node and allocating link bandwidth dynamic,we reduced the overall service time of multiple network services and completed the virtual network function scheduling.The main research are as follows:1.A method for service function chaining deployment based on dynamic programming is proposed.This method considers two types of delay which are processing delay and transmission delay,finally designed the deployment strategies with minimum delay cost of network services using dynamic programming.Simulation results show that the proposed method achieves better performances in terms of average processing time,request acceptance rate,gain and execution time.2.A method for virtual network function backup based on Backup-Cost Importance is proposed.First,by ordering the Backup-Cost Importance of the virtual network functions,we make a joint backup of virtual network function with the largest and second largest Backup-Cost Importance in each iteration,and obtain the optimal backup strategy through the corresponding selection and update model.Finally,the method is compared with other three methods,the proposed method has better performance in the backup cost,the amount of used physical node,the number of accepted service requests and backup resource utilization.3.A method for virtual network function scheduling based on hybrid genetic algorithm and tabu search is proposed.First,a dynamic bandwidth allocation policy of virtual links is established using an integer linear programming.Then,the transmission delay of data traffic in virtual links is introduced based on a traditional flexible job shop scheduling model,and the corresponding virtual network function scheduling model for 5G is established.Owing to the complexity of the scheduling problem,the scheduling algorithm based on hybrid genetic algorithm and tabu search is developed for solving the problem efficiently.The algorithm introduces tabu searches in optimization process of the genetic algorithm to balance capabilities of global and local searches,solves the problem of premature convergence of the genetic algorithm,and obtains better scheduling solutions.Simulation results show that the proposed algorithm outperforms the genetic algorithm in reducing the scheduling makespan effectively,and cater to 5G service with stringent delay requirements,thereby increasing users' experience and operators' revenues.