Research On Intelligent Resource Scheduling Strategy In Software Defined Network

With the widespread popularization and development of the Internet,networks become more complicated and difficult to manage,which seriously hinders the innovation and development of the network.In particular,the routers that carry the core functions of the network are expanding their capabilities and now integrate multiple functions to make them more and more complex and harder to control.For the time being,the main goal is to improve performance and extend functionality.The emergence of a new type network architecture-SDN solves the problems of being bloated and difficult to control TCP / IP structure of traditional networks.SDN separates network control from packet forwarding,providing logically centralized control and network programmable,simplifying network management and greatly improving network innovation and network performance.However,there are many problems with OpenFlow based on SDN technologies,such as the control plane scalability problem and the SDN forwarding plane design problem.Therefore,it studies the scalability of the control plane,based on the routing topology in the control plane,combining the characteristics of SDN control layer and forwarding layer,using OpenFlow as SDN implementation and Floodlight as controller and Mininet as simulation software to optimize the performance of delay,calculation time,transmission time,bandwidth,delay jitter and packet loss rate.Two optimization algorithms are proposed.The first one is the optimization of time delay and calculation time performance.Based on Yen's routing algorithm and A* algorithm,an efficient routing algorithm and FMRA algorithm are proposed to optimize the update speed of SDN routing topology.Simulation results show that with the increasement of the number of switches and the number of links,FMRA and Yen's delay of routing transmission tends to millisecond,and FMRA algorithm achieves minute-level optimization of topology update calculation time.When performing topology updates,FMRA reduces user latency significantly,improving network latency and reducing controller overhead.The second one is to optimize the transmission time,bandwidth,delay jitter and packet loss rate performance.Based on the characteristics of spatio-temporal data,a deep convolutional neural network algorithm is used to predict the link frequency of the network path.According to the prediction results,a link-load balancing routing algorithm and LLBR algorithm are designed to optimize the link load performance.The simulation results show that LLBR algorithm,RRS algorithm and WRRS algorithm are respectively used for route calculation and delivery in the same environment.By comparing four performance indexes of transmission time,bandwidth,jitter delay and packet loss rate,it is concluded that LLBR algorithm is superior to RRS algorithm and WRRS algorithm as a whole.