Research On Flow Table Lookup Performance Optimization Method For OpenFlow Switch

Compared with the closedness and complexity of traditional networks,Software Defined Networking(SDN)realizes the separation of network data plane and control plane,and provides flexible network programmability,which has become the fastest growing.One of the future development directions of the network has received extensive attention in recent years.OpenFlow is an SDN southbound interface protocol based on the concept of "flow".The SDN controller can use the OpenFlow protocol to deploy policies on the data plane switches according to the "flow" feature.However,due to its early development,when OpenFlow is deployed in the data plane,there are still some data forwarding performance bottlenecks.Therefore,based on the work of other scholars,based on the local characteristics of network traffic and the lookup characteristics of flow table,this thesis proposes a more efficient OpenFlow flow table lookup algorithm.The research work in this thesis is as follows:(1)For the flow table lookup problem of OpenFlow hardware switch,this thesis proposes a flow table distinguishing storage algorithm based on the local characteristics of network traffic.The scheme combines the local characteristics of network traffic,and designs the packet into a batch stream identification method.The network stream is divided into a packet batch stream and a packet decentralized stream,and combined with the difference of the storage medium,the packet is divided into a batch stream and a packet dispersed stream.In TCAM and SRAM.Finally,the algorithm proposed in this thesis is evaluated by means of network traffic samples.The experimental results show that the algorithm can effectively improve the TCAM hit rate in the hardware flow table architecture,speed up the OpenFlow flow table search rate,and effectively improve the OpenFlow storage and forwarding performance.(2)For mask traversal is the key factor affecting the performance of OpenFlow virtual switching.Based on the previous research points,this thesis designs the stream identifier and flow table entry pointer corresponding to the most recently matched connection by means of the local characteristics of network traffic.Stored in the flow table entry cache array,so that when the same batch of data packets arrives in a row,the mask array traversal is avoided and the corresponding flow entry is quickly matched.Finally,the algorithm proposed inthis thesis is evaluated by means of network traffic samples.The experimental results show that the proposed algorithm outperforms the current mainstream virtual flow table storage scheme OVS in cache hit ratio and average search length,which can effectively improve the virtual flow table lookup speed improves data forwarding performance of virtual flow tables.