Research On High Speed Communication In 5G Network Slicing

The access of a large number of intelligent devices and the emergence of various new business scenarios have posed new challenges to the current mobile networks in terms of network bandwidth,transmission delay,efficient control and energy saving.In order to meet the challenges of the 5G three major business scenarios and improve the quality of mobile network services,network slicing technology has emerged.Based on network function virtualization technology,the network function in slice network are implemented in form of software modules,and hosted by standard commercial virtual machines(VMs).Therefore,the high-speed communication technology among the virtual machines in the network directly determines the service quality of the 5G network.And the research about high-speed communication in 5G has become a hot topic.Aiming at the high-speed communication technology between virtual machines,this paper conducted researches from two aspects:the data forwarding among multiple virtual machines in a single physical server and the impact of virtual machines location on bandwidth utilization in a multi-physical server cluster network.The main innovations and research work of this paper are as follows:First,aiming at the problem of large data forwarding latency,low data throughput when using software switches in a single physical server for data forwarding among multiple virtual machines,this paper designs a hardware-based embedded packet forwarding system to optimize the packet identification process and realize efficient and flexible data forwarding.And the experimental platform verifies that,in the case of 440bits fuzzy matching sequence,the maximum forwarding delay is 1323.4 nanoseconds(corresponding to the maximum frame length of 1518 bytes)and the minimum forwarding delay is 68.0 nanoseconds(corresponding to a minimum frame length of 64 Bytes).Meanwhile,the corresponding data processing rate is 15.63Mpps(million packet per second)and 2.64Mpps.The performance is far superior to traditional software switches and DPDK-based software acceleration scheme.In addition,based on the system,this paper proposes an access scheme in 5G fronthaul network,which combines passive optical network with the hardware system.The experimental results show that its performance meets the requirement of 5G network delay.Second,aiming at bandwidth utilization optimization problems of dealing with the same task by different physical servers,this paper explores the relationship between virtual machine location planning and the redundant bandwidth reservation among multiple physical machines from the aspects of the network topology between physical servers,traffic model of three typical 5G business scenarios,internal resource capacity of server and physical resource requirements of business processing.After ing and modeling the problem based on the study above,this paper gives a constrained optimization model.In order to reduce the computational complexity and obtaining the optimal location scheme in controllable computing time,this paper applies genetic algorithm to solve the problem.Simulation results show that after the optimization by the algorithm,the redundant reserved bandwidth in the network is reduced by 98%,when compared with which before optimization.In addition,this paper also discusses the influence of algorithm parameters and network traffic cost model on optimization results.