Design of a Scalable Control Plane for Large-Scale SDNs
Software-Defined Networking (SDN) has become a major facilitator to flexible and programmable network operation; nevertheless, the extension of the control plane to handle big networks with heterogeneous networks has remained a core issue. The proposal presented in this paper is a design of a scalable control plane in a large SDN that can overcome the constraints of the centralized and fixed multi-controller designs. The suggested architecture takes the form of hierarchical and multi-controller structure with dynamically clustering and federation techniques to provide elasticity, fault tolerance, and effective use of resources. An assignment-based strategy to load-aware controllers is presented to dynamically assign switches to controllers in regard to real-time traffic intensity, control workload, and latency constraints whereas seamless controller migration ensures continuity of services in the face of network dynamics. The control plane also uses flow rule partitioning and aggregation to decrease state overhead and minimize redundant control operations in order to further ensure scalability. There is control message batching, adaptive poll, event based update and they are combined to reduce signaling overhead and enhance responsiveness by a large margin. A latency-aware controllerswitch mapping algorithm is also used to minimize control path delays to get rule installation and network convergence on time. Extensive performance analysis shows that the proposed design is almost linearly scalable as the network size is increased, and control latency is significantly lower and throughput is much higher than traditional SDN control plane designs. These findings affirm that the suggested scalable control plane is practical in supporting large scale SDN implementation and is therefore applicable in future carrier grade, data center and wide area network deployments at realistic workloads with varying topological setups in the modern programmable networks in the world.