Study On Optimization Technology Of Open Flow-table Based On Packet-flow Classification And Aggregation

The control and data plane are decoupled in Software-Defined Networking(SDN), which provide a new solution for researches of future Internet technologies. However, in OpenFlow-based SDN, the flow table’s scalability is poor and thus an OpenFlow switch can suffer from limited flow table size. This thesis studies on the optimization techniques of flow table.This thesis firstly discusses the technical characteristics of SDN and OpenFlow protocol specification. Then, the thesis gives a sumary of logical structure and features of a flow table. The size of flow table in an OpenFlow switch grows explosively due to the increasing number of fields and redundant structure of flow entries. In addition, the thesis analyses the compressible property and cost issue of flow table. To this end, the thesis proposes label-forwarding flow table optimization scheme and sub-flow table split optimization scheme.By focusing on sub-flow table splitting optimization scheme, the thesis elaborates the definitions of flow entry lookup and establishes an analysis model of flow table storage optimization problem. According to flow table split technical characteristics, the thesis forward flow table split algorithm based on packet-flow Classfication(CsplitAlg). Then, several numerical experiments are designed and carried out to validate the effectiveness of CsplitAlg and evaluate the Flow entry lookup effciency. The result, by dividing a single flow table into multiple sub-flow tables properly, shows that CsplitAlg is able to achieve the effect of saving flow table storage space at the expenses of flow entry lookup time.