High Performance Lookups Method Of Flow Table For OpenFlow Switch

The closeness in traditional networks greatly limits the flexibility and scalability of networks,which it will not be able to meet the current application trend of cloud computing,big data and server virtualization.In this background,SDN(Software Defined Networking)comes into being as a novel network paradigm.It segregates data forwarding from control logic,and facilitates switches to focus on the implementation of fast packet forwarding.When it comes to practical deployments,OpenFlow software switches are faced with heavy performance bottlenecks,where flow table lookup is a key step.Compared with the method of flow table lookups for OpenFlow switches by other scholars proposed,we propose more efficient lookups arithmetic by utilization of network traffic locality.The studies of this paper are as follows:(1)For OpenFlow software switches,this paper proposes an efficient lookup algorithm for virtual OpenFlow flow tables by applying a heuristic strategy.Firstly,we point out that mask traversals have a critical impact on the lookup performance of virtual flow tables.Then,the move-to-front heuristic is applied to self-adaptively adjust mask sequence by utilization of network traffic locality,which aims to accelerate the hit speed of mask lookups for subsequent packets.Finally,we evaluate the lookup performance of our proposed algorithm in virtue of real network traffic traces by experiments.The experimental results indicate that our proposed algorithm has a lower average search length than the flow table search methods in prevalent software switches,which speeds up flow table lookups and promotes the data forwarding of OpenFlow software switches effectively.(2)For OpenFlow hardware switches,this paper proposes a hierarchical storage and heuristic lookup scheme for flow tables(HHFT).We first employ TCAM combined with SRAM to hierarchically store large-scale flow tables.In particular,network flows are distinguished into elephant ones and mice ones in terms of network traffic locality,and their flow entries are stored respectively in TCAM and SRAM to get a high hit ratio of TCAM.As for SRAM with heavy lookup overheads to support wildcarding,we respectively apply the Move-Ahead-One(MA1)and Move-To-Front(MTF)heuristics to adaptively adjust the order of masks and flow entries for each successful lookup.Finally,we evaluate the lookup performance of our proposed flow table lookup scheme in virtue of real network traffic traces by experiments.The experimental results indicate that our proposed flow table scheme has a higher TCAM hit ratio and lower average search length than the traditional one with the speed-up ratio up to 2.5 or above.By this way,we effectively improve the lookup performance of flow table and promote the data forwarding of OpenFlow switches.