Research On Reinforcement Learning-based Service Function Chain Elasticity And Trusted Deployment Mechanism

With the rapid development of the Internet,the continuous increase of communication services have caused a continuous increase in network traffic,which poses a huge challenge to the problem of traditional network resource allocation.SDN and NFV technologies enhance the flexibility and scalability of the network.At the same time,network slicing can provide flexible resource allocation.In this process,VNF orchestration,deployment and scheduling are three very critical issues,which are pivotal for the performance of network services and the quality of user experience.At the same time,with the widespread application of network technology,higher requirements for communication quality and data privacy have been put forward,so the security of SFC becomes very important.Therefore,this thesis focuses on VNF deployment and orchestration and security,studies the optimization of virtual network resources based on reinforcement learning algorithms,and achieves the reasonable and efficient use of network resources through the efficient virtual network mapping mechanism and the trusted deployment orchestration mechanism of network slicing,and meets the diverse needs of user services.The main research contents are as follows:(1)In response to the diversified services requirements and the inability to flexibly allocate physical resources,a flexible deployment scheme of virtual network functions based on the Qlearning algorithm is proposed.The scheme ensures SFC can migrate the VNF of the overloaded physical node to other nodes or scale to satisfy the performance requirements of the SFC.Aiming to minimize the delay and energy consumption of the SFC,the VNF scaling problem is modeled as a nonlinear optimization problem,and the problem is mapped to a reinforcement learning process,in which the state space,action space,and reward function are defined as available CPU and link resources,migration or scaling decisions,and the negative correlation of the total costs overhead respectively.The simulation results show that the proposed algorithm reduces the delay in processing SFC while increasing the ratio of successful VNF scaling compared with the greedy-based algorithm and the random selection algorithm.(2)In response to the trust and self-adaptation problems faced by resource allocation in largescale heterogeneous network scenarios,a reliable and self-adapting SFC orchestration architecture based on blockchain technology and DDQN algorithm is proposed.Four stages are involved: smart contract for logging incontrolauthenticates the identity of the information requester;smart contract for SFC orchestration drives deep reinforcement learning for SFC orchestration;VNF deployment and traffic routing;smart contract for storing transactions in the blockchain stores the links between interconnection information,attributes and the sequence of the VNFs in the SFC in the blockchain.At the SFC orchestration stage,a DDQN-based SFC orchestration algorithm is proposed,with the goal of minimizing the cost of orchestration migration,while considering resource constraints such as nodes,links,and delays,establishes a corresponding optimization model,and use the DDQN algorithm to solve it.The simulation results show that the proposed algorithm has a faster convergence rate and greatly reduces the SFC orchestration delay compared with the DQN-based SFC orchestration algorithm.(3)Based on the Open Stack cloud platform,the VNF-FG service function chain orchestration was experimentally verified.A SFC including 2 virtual network functions was created,and the physical topology was built to complete the configuration of basic information such as flow tables,data flows and links.The experimental results show that,the network traffic can pass the corresponding VNF according to the predetermined orchestration after the service function chain is deployed,which proves the success of the SFC orchestration.