Research On Flow Entry Conflict And Update Consistency Problem In SDN

Traditional network architecture has become a bottleneck for future network development due to the lack of flexible adaptability to the requirements of differentiated service and resource virtualization of network.Software-defined network SDN solves this problem from an architectural level.It provides network-programmable capabilities by separating the control plane from the data plane.It enhances the capability and flexibility of rapid network evolution and reduces the cost of the network management.The overall efficiency of network is effectively improved.As a novel network architecture,SDN architecture leads the way of future networks.However,due to its ongoing technical advance,a few internal mechanisms are still imperfect.The flow entry management is a core technology in SDN.The consistency of its logic and update is essential to the network performance.When a logical conflict occurs,some SDN applications are not functioning properly,and the control plane logic is chaos or even invalid.When flow entry is updated inconsistently,black hole nodes or cyclic routes would appear,and the data plane would become transiently unreachable.Therefore,solving the consistency problem of flow entry logic and update has important research significance in the SDN architecture improvement.In this thesis,the conflict problem of flow entry and the consistency of flow entry update are solved through a B+ tree-based flow entry conflict resolution strategy and a flow entry update policy based on the guide packet respectively.The main contribution of the thesis are as follows:Through the in-depth study of the data structure of the flow table items,the feature extraction and modeling of the flow table items are completed.the mapping relationship between the data structure of flow table items and the geometric space is proved,which provides a theoretical basis for the type division and detection elimination of the flow table item conflict problem.According to the structural characteristics of the data,a B+ tree-based conflict detection algorithm is proposed,which avoids the backtracking problem caused by the Trie Tree algorithm for collision detection.Compared with the Btree collision detection algorithm,multiple collisions can be detected with a low cost.The delivery process of the flow entry is modeled and formalized.And a flow entry update strategy based on the boot data packet is proposed.This strategy guarantees the order of flow entry updating by converting the unordered flow entry delivery process into an ordered packet forwarding process in the data plane.Theoretical proof shows that this strategy can guarantee the consistency of updating at package consistency level.Compared with the existing update strategy,it can effectively reduce processing cost.A simulation experiment platform is built to verify the conflict resolution strategy of flow table items.This strategy is comprehensively evaluated in three indicators: effectiveness,time overhead and controller load.According to the experimental results,the applicable scale of the strategy is analyzed.The strategy maintains a linear growth under single controller condition,and the experiment performs well.So this strategy shows some practical values.