Hierarchical Edge-Cloud SDN Controller System With Optimal Adaptive Resource Allocation for Load-Balancing

Abstract

Although the openflow-based software defined network (SDN) architecture can ease the workload of network control and management and separate it from the switch/routing operations, a computation-resource limited controller can still be congested by heavy flows and then experiences serious delay. To enhance network scalability and reduce computation delay on SDN networks under Quality of Service (QoS) requirements, a hierarchical edge-cloud SDN (HECSDN) controller system design is proposed with three features. First, by sharing computational resources in the edge and the cloud, the system architecture provides a flexible mechanism for devices to allocate their computational tasks according to traffic loads. The second feature is to design a queueing model of the proposed architecture. The model description of the networking architecture enables the network designers to quickly estimate the performance of design without considerable time and cost in experimental setups. Third, derived from the queueing model, an efficient load-balancing algorithm satisfying QoS requirements or fairness allocation for different applications in the HECSDN architecture is proposed. This newly-developed multi-tier controller system has been proved to be effective even when working on a large-scale SDN, without sacrificing the overall performance. Moreover, this system stays highly stable in transient periods even under highly fluctuating flow arrivals.