Joint Resource Optimization Of Control And Data Planes In Software Defined Networks

Software defined networking(SDBN)separates data plane and control plane,and provides global view for the control plane to make efficient decisions and network management to guarantee the quality of service(QoS).Limited resources(e.g.,TCAM,computing capacity and link bandwidth)on control plane and data plane in SDNs should be optimized to handle increasing network traffic and offer diverse services.Moreover,network uncertainties,including traffic dynamics and topology asymmetry,bring chal-lenges to current resource optimization method as well.Though some solutions(e.g.,ECMP)satisfy resource constraints,these methods do not work well especially for traf-fic dynamics and network asymmetry.Moreover,other solutions achieve load balanc-ing with additional hardware or software resources,which increase the system cost and limit the applicability.Therefore,it is necessary to optimize network resources while conquering network uncertainties.For the single controller and multiple controllers in SDNs,this paper separately designs resource optimization methods.For the former one,this paper designs PrePass for load balancing to optimize computing resource on the control plane and resources on data plane,which installs wildcard entries for some aggregate flows in advance to satisfy the flow table size constraint and reactively install flow entries for the remain-ing flo'ws.This paper also shows PrePass can handle network uncertainties.For an SDN with multiple controllers,this paper designs RDMAR algorithm to balance loads of both links and controllers,which can optimize both resources on data and control planes.This paper formulate both problems and prove their NP-hardness.This paper also presents rounding-based algorithms and analyzes that the proposed algorithms can achieve constant bi-criteria approximation under most practical situations.Thei paper implements PrePass on a real SDN testbed and conducts large-scale simulations for both problems.The experimental results and extensive simulation results for PrePass show that the proposed method can satisfy different resource constraints on switches,and only increase the link load ratio by about 5%-10%compared with per-flow routing scheme under various traffic scenarios.Likewise,simulation results show that the RD-MAR algorithm can reduce the maximum controller response time by 70%compared with the previous solution,while only increasing the maximum link load by 3%.