Research On Intelligent Orchestration Algorithm Of Service Function Chain In Network Function Virtualization

In the future,the network will support diversified business scenarios,where the number of terminals will also increase substantially.The existing network architecture is difficult to provide strong support for it.Network Function Virtualization(NFV)technology breaks the current vertically rigid operating mode of the network,decouples network functions from dedicated hardware devices,and takes the form of Service Function Chain(SFC)as a customized service for users,which improves the flexibility and scalability of services and reduces network costs.SFC is a logical link that consists of a set of virtual network functions(VNFs)in a particular order.How to orchestrate SFC in the underlying physical network is a key issue in NFV.This thesis divides the SFC orchestration problem into SFC deployment and reconfiguration,and has a detailed research on it.The main innovative contents are as follows:Aiming at the SFC request of wireless users,an SFC deployment and resource management scheme with wireless resource allocation is proposed.First,an environment-aware SFC resource allocation mechanism is constructed,and the SFC deployment cost minimization model with the constraints of delay requirements,wireless rate requirements,and resource capacity is established;further,the optimization problem is transformed into a discrete-time Markov Decision Processes(MDP),and considering the high dimensionality of the MDP state space and action space,a SFC deployment and resource allocation algorithm based on deep deterministic policy gradient(DDPG)is proposed to solve the MDP.Simulation results indicate that the algorithm can effectively reduce SFC deployment costs and end-to-end latency,while meeting the SFC performance requirements and resource capacity constraints.After the SFC deployment is completed,the wireless rate requirements of the SFC may change,which may cause the load of server where VNF instances(VNFIs)are deployed to be too low or too high.In order to solve this problem,a SFC reconfiguration algorithm based on Actor-Dueling-Critic(ADC)reinforcement learning is proposed.SFC reconfiguration includes the migration of VNFIs and resource reallocation.According to the environment-aware SFC resource allocation mechanism constructed in this thesis,the process of resource demand changes is explained firstly,and a VNFI migration overhead minimization model with the constraints of server loadbalancing and SFC performance requirements is established;further,the above optimization problem is transformed into an MDP model,and the ADC reinforcement learning algorithm is leveraged to obtain the SFC reconfiguration strategy.Simulation results indicate that this algorithm can effectively reduce the VNFI migration overhead and save the server resource consumption to a certain degree,while ensuring the server load balancing and SFC performance requirements.