Research On Modeling And Algorithm For Virtual Network Functions Mapping In Elastic Optical Networks

In recent years,with the rapid development of internet technology,network traffic jam happen much easier and more frequently.The traditional wavelength division multiplexing networks with fixed channel spacing and modulation format cannot suit the increasing traffic demand in the network.The elastic optical network can provide the flexible and variable bandwidth allocation to each connection request and get higher spectrum utilization.Also,network function virtualization technology separates network functions from the underlying physical devices to form virtual network functions,so as to deploy them on the highperformance datacenter,which significantly saving operation cost for network service providers and motivating deployment of new network functions.However,how to properly deploy virtual network functions in the elastic optical network to minimize cost is a challenging problem.This thesis mainly studies virtual network function mapping problem considering datacenter allocation and virtual network function mapping in multicast routing.The main contributions of this thesis are as follows:1.The virtual network function mapping considering datacenter allocation is investigated.High-performance servers in datacenters are expensive resource,and the number and location of datacenters affect the mapping scheme of virtual network functions.Therefore,we establish a virtual network function mapping model considering datacenter allocation and whose objective is to minimize the network resource cost(datacenter resource cost,virtual network function resource cost,spectrum resource cost).To solve this model,an improved genetic algorithm is used in the phase of datacenter allocation,including a new crossover mutation operator and a local search operator for speeding up the convergence of the algorithm.The process of virtual network functions mapping to high-performance servers in datacenters and routing adopt a new load balancing strategy.For spectrum assignment,the first fit strategy is used.In order to verify the validity of the proposed model and algorithm,a series of experiments are conducted in different network topologies.The experimental results show that the proposed model is reasonable,and the algorithm is effective.2.The multicast-related virtual network function mapping problem is investigated.The problem of multicast routing and spectrum allocation is very complex.This problem will be very challenging when trying to jointly optimize the location of virtual network functions,multicast routing,and spectrum allocation.To solve this model,a virtual network function mapping model in multicast routing is established and whose objective is to minimize the network resource cost(virtual network function resource cost,the total cost of generating frequency slots,maximum index of used frequency slots).In order to solve the model effectively,we encode the routing scheme and the virtual network function mapping scheme and the two-population co-evolutionary algorithm is proposed to solve the problem.A uniform population initialization method for generating better initial solutions and a crossover mutation operator with strong search ability is designed.In order to verify the effectiveness of the proposed algorithm,a series of experiments are conducted in different network topologies.The experimental results indicate the effectiveness of the proposed model and algorithm.