A 100Gig network processor platform for openflow

Openflow splits the control plane from the data plane and move it to a centralized remote controller that dictates the forwarding behaviour to multiple Openflow switches. OpenFlow 1.1, introduces pipeline processing of the flow tables. This helps to improve flexibility and performance. This paper presents an implementation of the version 1.1 of OpenFlow capable to achieve a good performance. Knowing that OpenFlow is a rule-based approach, we can have several levels of flexibility of the switch by only updating the flow entries on-the-fly. We will show also that we can achieve a good performance using RFC (Recursive Flow Classification) that uses SRAM and TCAM side by side to enhance classification performance for the Openflow 1.1 Switch. Using this approach helps ameliorating the parallelism of the classification process and better exploiting of the hardware resources since most hardware platforms (such as Network processors) contain both TCAM and SRAM to store the lookup structures.