SDN Flow Table Update Method Based On Consistency Of Application Layer Policy Conversion

With the continuous development of Internet, the increasingly complex network applications and management requirements prompt us to seek a more efficient network architecture, and Software Defined Network(SDN)comes into being. SDN could separate control and forwarding to change the existing network architecture. The users can not only control and operate the underlying network in a global perspective, but also deploy the application layer policy to the underlying network according to their requirements. Because SDN originated from Openflow, the design and the use of flow table is the core of Openflow, SDN supports the flexible application layer policy through the timely update of flow table, but how to keep the consistency of flow table update has been a key research of SDN. The thesis analyzes three different schemes of application layer policy deployment, and puts forward a scheme of application layer policy conversion to solve the generic problem of the controllers and the efficient deployment issue of the policies theoretically. Then the thesis further expands the scheme of application layer policy conversion to study the merge issue and update consistency problem of flow tables. For the flow table merge issue, the thesis designs a two-stage flow table merge algorithm which is also helpful to solve the insufficient storage space problem of underlying switches. The algorithm merges the similar flow tables converted from the same policy or different policies to reduce the number of flow tables. For the flow table update consistency problem, the thesis presents a flow table update consistency scheme. The merged flow tables can be updated at different time periods according to the respective policy and packets can be processed in their network transport lifecycle only in accordance with the new or old flow tables which are converted by a policy. The scheme can ensure the update consistency of flow tables converted from different policies, and ultimately ensure the overall consistency of the application layer policy conversion. Simulation results show that the two-stage flow table merge algorithm is applicable to the three experimental scenes: first stage flow table merge scene, second stage flow table merge scene and two-stage flow table merge scene, and could solve the merge issue of flow tables. What’s more, the flow table update consistency scheme is suitable for two experimental scenes: the changed forwarding path scene and the changed forwarding port scene. And to some extent, it solves the update consistency problem of flow tables.