Research On SDN-based Multi Dimensional Resource Allocation Strategy Driven By Service

With the development of 5G network,the future 5G network will provide network service support for diversified business scenarios,which have differentiated service characteristics and service quality requirements.At the same time,with the rapid development of the emerging vertical industry business,the demand for multi-dimensional resources such as communication,computing,and caching for services has exploded.Therefore,it has become an important issue to allocate multi-dimensional resources for differentiated services,ensure diversified service quality requirements and improve resource utilization efficiency.Network slicing technology can divide a physical network into several virtual networks,and allocate differentiated resources to each virtual network for a specific type of application.The Software Defined Network(SDN)technology can monitor and allocate the global resources of the Network due to its characteristics such as control and data plane separation.Therefore,SDN technology and network slicing technology have become key technologies in 5G network and are used to provide differentiated resource allocation for business scenarios with different requirements.An important task of network slicing is to provide the optimal allocation of resources(such as bandwidth,hardware,memory,program,etc.).Reasonable allocation of resources can improve the utilization of network resources,reduce the cost of the entire network,improve the quality of service,and meet the needs of users.In this thesis,the network slice architecture and SDN technology are combined to carry out service type identification and communication-computing multi-dimensional resource allocation for various business scenarios and applications.Based on network slice architecture and SDN technology,a resource allocation scheme for intelligent service identification is proposed.First,this thesis proposes a business type recognition method based on neural network,using the method of neural network according to the statistical features of the packet flow data is used to identify the type of business,business type and traffic data of accurate identification,and then slice allocation differentiation for different business types of network communication-multidimensional resource calculation.At the same time,based on the technology of SDN structures,experimental platform,the traffic data collected production data set used for neural network training and testing,verify the feasibility of resource allocation scheme,discussed using the different methods used to identify the type of business to business recognition results,the influence of intelligent business to identify and analyze the resource allocation effect on the performance of throughput and user satisfaction.Furthermore,a multi-dimensional resource allocation method based on reinforcement learning is proposed based on the above resource allocation scheme.The allocation of communication-computing resources is further optimized.Reinforcement learning method is used to allocate differentiated communication and computing resources to network slices of different service types according to network state and resource situation,so as to improve spectrum efficiency and maximize system utility while ensuring the Quality of Experience(QoE)of users.The simulation is verified based on TensorFlow platform.The results show that the multi-dimensional resource allocation method proposed in this thesis can not only meet the requirements of differentiated service quality,but also improve the spectral efficiency.