Font Size: a A A

VNF Service Chain Placement In IP Over EON Networks

Posted on:2020-08-06Degree:MasterType:Thesis
Country:ChinaCandidate:Z J YangFull Text:PDF
GTID:2428330590478612Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
With the advent of network function virtualization technology,it is proposed to use software-implemented virtual network functions(VNF)to replace the hardware middleboxes used in the network.The flexible deployment of the VNF combined with the elastic optical network can significantly reduce the operating expenses of the network.In this paper,we study the deployment of VNF service chain in elastic optical networks considering the virtual network function service chain as a connection request.Firstly,this paper introduces the research background and research status of elastic optical network and virtual network function service chain deployment,and then expounds the key technologies of elastic optical network and network function virtualization.In the elastic optical network with limited capacity,we proposes an integer linear programming(ILP)algorithm to maximize network throughput for distance-adaptive modulation formats VNF service chain deployment.The simulation results show that the throughput of the VNF service chain deployment uses distance-adaptive modulation formats is better than the VNF service chain deployment uses single modulation formats.In an elastic optical network with sufficient resources,we optimize network energy consumption in a static traffic scenario.We compare the VNF-based service chain deployment with the hardware middlebox based service chain deployment based on ILP algorithms.The simulation results show that the VNF-based service chain deployment achieves lower energy consumption.Finally,considering the high complexity of the ILP algorithm,a reinforcement learning method(RL)based on the deep reinforcement learning DQN method is proposed.The simulation results show that the performance of the RL method can reach 90% of the ILP in six-node network topology.We also compared the performance of the RL method with the heuristic algorithm NFF(Nearest Service Function first)in the 14-node NSFNET network.The simulation results show that the energy consumption of the RL is lower than the energy consumption of the NFF,and as the number of services increases,the advantage of RL in energy consumption increases.
Keywords/Search Tags:IP over EON, Virtual network function, Service chain, Reinforcement learning
PDF Full Text Request
Related items