Research On Fast And Consistent Path Reconfiguration Strategy In Software-defined Network

Software-defined networking provides a framework for dynamically adjusting and reconfiguring the data plane using flow rules,and it is also one of the key technologies for realizing network service customization.The key to realizing the high dynamics of software-defined networks lies in the ability of the network to respond to faults or traffic changes in the network in a very short time,which leads to frequent path configuration update operations in the software-defined network.However,although the software-defined network is logically centralized,the system composed of switches located in the data plane is still distributed.This leads to inconsistent states such as black holes,loops,congestion,and deadlocks during the configuration update process of the initial path and the target path.These inconsistent configuration updates will cause network interruption,seriously affecting the overall network performance and user experience.In addition,if the update process takes too long,it may cause the target configuration to become invalid.Therefore,on the basis of ensuring the consistency of the path configuration in the software-defined network,reducing the time of path reconfiguration is a very important and urgent problem to be solved.At present,most of the existing researches on network path configuration update focus on the conversion process between the current initial path configuration and the target path configuration.Researchers propose different schemes to ensure consistency in the configuration update process.However,some of the existing work can only guarantee the requirement of weak consistency in the update process of initial and target configurations,and some solutions perform poorly in the update efficiency of the configurations,resulting in a longer time consuming update process,which affects the user experience.In addition,most of the existing work ignores the influence of different target configurations given by the controller on the path reconfiguration time.Therefore,the main research content of this thesis is to reduce the path reconfiguration time by combining the selection of the target path configuration of the control plane and the improvement of the data plane update efficiency on the basis of ensuring the consistency of the data plane forwarding rule update process,so as to realize the fast and consistent path reconfiguration in software-defined network.The main research content and contribution points of this thesis are as follows:(1)We consider the effect of the target configuration selected by the controller on the network path reconfiguration time.First,we analyze the time model during network reconfiguration and find that the number and type of rules that need to be updated by the switch during reconfiguration are different when different target path configurations are chosen.Then to minimize the rule deployment time on the switch for each flow during path reconfiguration,we formulated the shortest configuration path problem.To solve this problem,we propose the shortest deployment time path selection algorithm,which selects the optimal configuration as our updated target path configuration based on the type and number of rules that need to be deployed on the switch during the configuration update.Finally,simulation results show that our proposed shortest deployment time path algorithm can significantly reduce the time to deploy rules on the switch during the path configuration transition compared with the current work.(2)We design a dynamic algorithm for updating forwarding rules on the data plane that balances consistency and efficiency in response to the low efficiency of forwarding rule updates.The algorithm guarantees the property of no black hole and no loop in the update process by means of reverse-order update,improves the update efficiency by segmenting the current path and the target path to update in parallel,and guarantees the congestion-free update process by constructing the dependency graph,and solves the possible deadlock problem by means of speed reduction.Finally,the simulation results show that our proposed the Consistent Schedule Update algorithm can effectively reduce the update time compared to existing algorithms and guarantee the consistency property during the update process.