Study On Function Node Deployment And Flow Routing In Virtual Network

With the increasingly widespread application of the Internet,the security of data transmission,resource utilization,and system performance have gradually become the focus of attention.Software-defined networking(SDN)changes the traditional network architecture,deconstructing it into two planes of data and control,so that operators can manage the network effectively.At the same time,the implementation of many network functions depends on hardware device.This will cause a surge in network cost and increase the complexity of network configuration,thereby reducing the quality of service for network operators.As a result,Network Function Virtualization(NFV),a new network technology,has received widespread attention from industry and academia.NFV implements network functions by running network software on virtual machine(VM)through virtualization technology.Due to the flexibility of virtual machine,a network containing NFV can provide a low-cost,easy-configuration,high-quality network service environment.The main research content of this paper is the deployment of virtual network function nodes and data flow routing in the data center network under the SDN architecture.First,this thesis proposes a network framework containing virtual network function nodes,using the idea of ?service chain to meet the network functions on each data flows are processed in a specific order.From the perspective of the network service provider,we should minimize the configuration and transmission cost,and from the perspective of users,we should improve the quality of service and reduce delay.Finally,this thesis proposes a dynamic deployment algorithm of virtual network function nodes based on online learning to meet the energy-efficient data transmission between networks.The main research and results of this thesis are described as follow:1.Study the minimization of network operating cost of data center network with background traffic interference.Firstly,the operating cost is subdivided into three parts:deployment,transmission and processing to characterize the cost function.It is pointed out that avoiding the frequent changes of network topology is the key factor to reduce cost.Therefore,a prediction algorithm based on fast Fourier transform is proposed,and the prediction result is used to determine whether it is necessary to update the network topology.Secondly,in order to solve the problem of link preemption by background traffic,this thesis proposes a Upper Confidence Bounds(UCB)algorithm based on online learning.Simulation results show that compared with the traditional virtual network function node deployment algorithm,the proposed algorithm can effectively solve the problem of link congestion caused by background traffic,and greatly save network overhead.2.Study the optimal routing of data flow in a network containing virtual network functional nodes.We model the transmission of data flow under the mechanism of service chain and construct a transmission cost minimization problem under the constraint of maximum link bandwidth and maximum link delay.In particular,this paper constructs a multi-layer acyclic graph to satisfy the sequential nature of network function processing in the data flow.Simulation results show that the optimization algorithm proposed in this thesis is superior to conventional routing algorithms in different network topologies,which is conducive to its application in practical systems.In addition,the effects of maximum end-to-end delay and service chain length on network loss are also verified.