Research On Resource Manage Ment Technology Of Software Defined Passive Optical Network Slice

With the rapid development of the 5th generation mobile communication technology,new vertical industries and application scenarios are derived.The diversified service demands bring new opportunities and challenges to passive optical access networks.The passive optical network adopts a tree topology to meet users’ bandwidth requirements through bandwidth resource sharing mechanism.However,in the face of the differentiation of user demand and uneven distribution of load,how to ensure resource sharing and meet diversified user demand is the main problem that access network faces.Network slicing technology provides a solution to the above problems.The technology can run multiple independent customized virtual networks on the general physical infrastructure,break the limitation of the traditional unitary network structure,and provide differentiated service quality guarantee for users.In this thesis,the slice resource management technology of PON based on software definition is studied.Firstly,to meet the QoS requirements of time-sensitive traffic in the business-service slice of eMBB,a computing resource allocation algorithm based on time-delay optimization is proposed.Then,aiming at the problem of low bandwidth resource utilization between slices under the condition of unbalanced traffic demand,a network slice bandwidth resource allocation scheme based on traffic prediction is proposed.Finally,aiming at the problem of slice QoS congestion caused by ONU service demand fluctuation in ONU granularity slice,a dynamic Network slice algorithm based on three-way decision was proposed.The specific contributions of this paper are as follows:(1)Aiming at the problem that the QoS quality of existing optical access network for time-sensitive services in eMBB service cannot be guaranteed,this thesis proposes a cloud-side collaborative TDM-PON network slice architecture.On this basis,an algorithm of computing resource management based on time-delay optimization is proposed.Based on software definition and network function virtualization technology,the proposed network slice architecture is compatible with cloud computing resources and edge computing resources,and improves the flexibility of network slice resource allocation and network management.The proposed computing resource management algorithm enables delay-sensitive services of eMBB slices to dynamically share edge computing resources with uRLLC services under strict priority scheduling.The simulation results show that,the proposed algorithm greatly improves the utilization of computing resources,reduces the transmission delay of traffic,guarantees the service demand of eMBB slicing services with low delay,and improves the QoE of users.(2)Aiming at the problem of low utilization of passive optical network slice bandwidth resources caused by static bandwidth allocation algorithm,this thesis proposes a TDM-PON slice bandwidth allocation scheme based on LSTM neural network traffic prediction.In this scheme,the bandwidth demand is determined according to the predicted network slice traffic,and the bandwidth is allocated according to the load ratio to ensure the fairness of bandwidth allocation.In this paper,slicing is constructed with MFH and IoT scenarios as slicing application background.Simulation results show that this scheme can guarantee the requirement of MFH and IoT with low latency and effectively improve the utilization rate of slicing resources.At the same time,the efficiency and flexibility of resource allocation are enhanced,network delay and packet loss rate are reduced,network throughput is improved,QoS requirements of different priority slices are guaranteed,and operators’ capital expenditure and operating expenses are reduced.(3)Aiming at the problem that the ONU granularity TDM-PON slicing resource pre-planning scheme cannot guarantee the QoS of ONU users in the case of traffic burst,this thesis proposes a dynamic network slicing algorithm based on three-way decision model under ONU granularity.Based on the three-way decision theory of ONU service compliance rate,the algorithm quantifies the necessity of ONU migration.According to the decision result of service compliance rate,ONU is selectively migrated.Simulation results show that this algorithm can effectively reduce the delay and packet loss rate of ONU with high load,improve bandwidth resource utilization and improve users’ service experience under congestion state.