Research On Efficient Lookup Method Of OpenFlow Virtual Flow Table Based On Network Traffic Locality

With the popularization of communication technology in the world,the rapid growth of network access devices and user connections has brought a lot of troubles and inconveniences to today's network.However,the traditional network lacks flexibility and expansibility,and it is difficult to adapt to the changes of the network.In this background,Software Defined Networking(SDN)emerges at the historic moment.OpenFlow virtual switch is responsible for data processing in the SDN data plane layer.Due to its great advantages in cost and flexibility,it is gradually deployed to the network test platform.However,the introduction of wildcards in OpenFlow makes it impossible for the virtual switch in the SDN data plane to directly use hashing and other traditional efficient lookup methods,resulting in serious performance bottlenecks in the virtual flow table lookup process of OpenFlow.The OpenFlow virtual switch is implemented by tuple space search algorithm(TSS).As tuple space search requires sequential mask detection before successful search,TSS will incur a large amount of search overhead when there are many tuples.Therefore,speeding up tuple space search is the key to fast flow table lookup.The search process of convection table is explored and optimized in this paper.The research work is as follows:(1)Aiming at the problem of low efficiency of cache lookup of OpenFlow virtual switch,this paper designs an active and accurate flow cache scheme based on Cuckoo hash by combining network traffic locality and cleverly taking advantage of the characteristics of high utilization rate and low hash conflict of Cuckoo hash.The scheme bypasses the mask detection process and reduces the processing time of the switch by separating the active precise flows into the cache and finding the flow table entries directly according to the signature value and address in the cache during a lookup of the same flow(2)Considering the high cost of mask detection in OpenFlow virtual switch,this paper proposes a mask adjustment scheme with real-time characteristics by combining the network traffic locality and the mask ordering mechanism.In this scheme,the order of mask lookup is adjusted according to the principle of accumulation of more than one priority,so as to find the flow table items with fewer times of mask detection,so as to improve the packet classification performance of OpenFlow software.