Research On Resource Management Algorithms For SDN-Based Passive Optical Networks

The continuous development of society has led to higher and higher requirements for information technology.With the continuous emergence of high bandwidth requirements and diversified services such as cloud computing and ultra-high definition video,people are also demanding higher transmission capacity and speed of communication systems.As the last station to connect MAN,access network has gradually become the bottleneck of users'diversified business needs.The traditional access network has poor flexibility,poor scalability and weak network resource scheduling ability,which has been unable to meet the requirements of diversification and high quality of network services.In order to meet the new requirements of access network brought by social development,people have proposed to apply software-defined network to access network in recent years,so as to improve the flexibility of access network,and can use the advantages of software-defined network to collect global network information in real time and improve the efficiency of network utilization.In order to further improve network performance,people model network traffic and use traffic prediction algorithm.Aiming at improving the efficiency of network use,aiming at the current demand for high bandwidth,high quality and diversification of network,this paper mainly studies the network architecture and bandwidth resource management algorithm.A traffic model of access network upstream network is built on MATLAB simulation platform,and the simulation results are analyzed.The main research results of this paper are as follows:(1)A soft definition time division multiplexing passive optical network(TDM-PON)architecture is proposed to support the resource allocation algorithm in this paper.The architecture divides the network into two layers,control plane and forwarding plane,and divides the terminals of forwarding plane into several scenarios according to different traffic attributes.Through refinement of traffic classification,network performance can be improved.(2)A GRNN network model suitable for upstream traffic access network scenarios is proposed.Firstly,according to the mathematical principle of traffic characteristics,the scenario of upstream traffic access network is simulated.After that,the simulation proves the predictive ability of the model and the good predictive performance of GRNN network model.(3)Based on the above software definition of TDM-PON architecture and GRNN network model suitable for access network traffic prediction,a dynamic resource allocation strategy based on GRNN network prediction is proposed.The algorithm can collect the information of the whole network,and then take into account the real-time load of the network,to achieve more accurate scheduling,and according to different load conditions,to achieve adaptive bandwidth resource scheduling for different services.The simulation results show that the average packet delay of the network is reduced by 70%and the bandwidth utilization of the network is increased by 19%.(4)A dynamic resource management strategy based on virtual memory allocation is proposed.The algorithm can learn the global state in real time,fine allocation of the ONU buffer pool service cache value,and adaptive allocation of bandwidth resources for different services according to different load states.The simulation results show that the average packet delay decreases by 36%and the bandwidth utilization increases by 18%in the whole network using this algorithm.