Research For Flow Table Optimization Technique Of OpenFlow

Software Defined Network(SDN) implements the flexible configuration of network devices and the free allocation of data flow by decoupling the data plane and control plane network devices, also it allows users to control network behavior by using the open programmatically. However, SDN has the problem that flow table do not update timely when the network traffic at peak. Also, SDN has the problem of introducing network delay when compressing space for multiple-table.A timeout control scheme in SDN using predictive and load-aware optimization algorithm is proposed to address the problem that flow table do not update timely. Firstly, this algorithm collects the various newly arrived flows during per unit of time. Based the information, this algorithm estimates the number of newly arrived flows in the next time unit by using the second moving average method. Finally, this algorithm dynamically adjust the number of the flows in the table based on the dynamic load factor. So as to guarantee the flow table entry number, simultaneously, the interaction between the controller and switch is reduced and the controller workload is effectively lowered based on the above method. Experimental result show that the flow table matching rate and data forwarding rate are improved by using the proposed algorithm, the algorithm also increases the number of activities in the flow table.To effectively solve the problem of introducing network delay when compressing space for multiple-table, an adaptive method for multiple-table building and searching is proposed. Fistly, this method divides the flow table space into realtime update area and multiple-table area. Then a constrained inequalities group of multiple-table series is put forward based on studying the balance of compression rate of multiple-table and network delay. Finally, the single flow table is divided into multiple-table basing on the constrained inequalities group and the repetitive rate of match fields, implementing the robustness of compression and quickly search. The experimental results indicate that the method is not only robust to compress the flow table, saving the storage space, but also reduced the network latency introduced by the multiple-table compression, increasing the amount of data forwarding per unit time.