The Deployment And Optimization Of NFV Service Function Chain Based On Artificial Intelligence

With the rapid development of Internet and the upgrading of hardware equipment,network users demand is growing,the operator is still mainly rely on the erection of a large number of middle box to realize the user's network request and provide service for the user.However,due to the single function of middle box,and the constraints of its geographical location,this mode is more and more difficult to meet user needs.Network function virtualization technology offers a new way for the operators,this technology use software to realize the functions that originally based on proprietary middleware equipment,it translates network request to orderly arrangement in the number of virtual network function,only need to deploy it in a standard x86 devices.This way not only gets rid of the influence of the traditional middleware equipment location constraints,also save huge cost and maintenance cost of equipment,and compared with the existing system,it also has a great improvement in terms of performance.Artificial Intelligence technology has gradually come into our lives,and it is also becoming to promote the development of all walks of life,the problem that is difficult to solve before also began to be solved by using the method of artificial intelligence.Artificial neural network is a method of artificial intelligence,build a set of nonlinear signal processing system by imitating the biological nervous system working principle,and solve some issues with large-scale complex,not only on the time cost has a large optimization,but also has a good ability of parallel processing.Graph neural network is a kind of neural network which is particularly suitable for processing graph problems.Since the date of its introduction,it has been applied to various graph related problems.The working principle of the neural network is based on the state of the node,and the target output is generated through the analysis of the properties and state of the node.The contribution of the thesis is as follows:(1)In the case of NFV,a heuristic algorithm is always used to deploy the service function chain with different strategies for different indexes.However,the scope of the heuristic algorithm is easy to fall into local optimum,and the algorithm also can't reach the level of outstanding performance,the most important thing is that its computation time is relatively long,these shortcomings when dealing with large-scale network topology will be a greater degree of amplification.Therefore,this thesis uses the method of graph neural network to train the system according to the information in the network topology,so that it can deploy the service function chain autonomously.(2)According to the waste of nodes resources and bandwidth resources in the service chain function deployment problem,using the same method of graph neural network,puts forward an integration and split algorithm,split the nodes and integrate service chain integration,improve the resource utilization of nodes and bandwidth.