Research On Flow Table Optimization Strategy Under SDN Network Architecture

The SDN network is a new type of network that decouples the data plane and the control plane,and realizes the separation of hardware devices and logical functions through programmable performance.The switch uses flow entries to achieve efficient matching and forwarding of data packets,which provides a guarantee for the controller to centrally manage the network.At present,the switch flow table entry storage uses a ternary content addressable memory(TCAM),which is limited by the cost,hardware technology and other issues,and the capacity of the flow table is limited.Therefore,the optimization strategy for flow table storage is an important direction.In this thesis,a dynamic timeout mechanism based on flow classification and a nearest idle switch sharing algorithm are proposed for the optimization of SDN network switch flow table storage performance.The main ideas are as follows:1.This thesis proposes a flow classification-based dynamic timeout algorithm(Flow Classification – Dynamic Timeout,FC-DT).The core idea is to detect and classify the network data flow into elephant flow and mouse flow according to the occupancy rate of the data flow link,and limit the classification task to the first switch after the flow is started,so as to avoid the repetition of switches before the classification is completed.The monitoring causes the load of the controller to increase;according to the classification detection results,different timeout allocation strategies are adopted for the elephant flow and the mouse flow;that is,the elephant flow is given a long enough survival time,and the switch will intelligently adjust the size after each match.The actual idle timeout time of the image flow;at the same time,the invalid survival time of the mouse flow flow table entry is reduced as much as possible to release more flow table space.Finally,through the network simulation experiment,it is proved that this strategy can effectively reduce the data packet loss rate and reduce the load of the controller.2.Based on the analysis of data traffic,it is found that the SDN network will inevitably overflow at a certain time,which will lead to the degradation of network performance.Therefore,this thesis proposes a Nearest Idle Switch Transfer(NIST)algorithm.The core idea is to first reserve some space in the switch for the general matching flow table entry,and obtain the general matching flow table entry of the nearest idle switch to the switch by obtaining the network status of the controller;when a switch is found to have burst traffic When overflow occurs,the generated Table-Miss message is not uploaded to the controller,but uses the general matching flow entry to transfer the data packet to the idle switch and reapplies for the flow entry.Through experimental comparison,it is found that the algorithm can effectively alleviate the overflow of data packets,and will not significantly increase the forwarding bandwidth occupation.