CentFlow: Centrality-Based Flow Balancing and Traffic Distribution for Higher Network Utilization

Abstract

Next-generation networks (NGNs) are embracing two key principles of software defined networking (SDN) paradigm functional segregation of control and forwarding plane, and logical centralization of the control plane. A centralized control enhances the network management significantly by regulating the traffic distribution dynamically and effectively. An eagle-eye view of the entire topology opens up the opportunity for an SDN controller to refine the routing. Optimizing the network utilization in terms of throughput is majorly dependent on the routing decisions. Open Shortest Path First (OSPF) and Intermediate System to Intermediate System (IS-IS) are well-known traditional link state routing protocols proven with operation over operator networks for a long time. However, these classical protocols deployed distributively fall short of expectation in addressing the current routing issues due to the lack of a holistic view of the network topology and situation, whereas handling enormous traffic and user quality of experience (QoE) requirements are getting critical. IP routing in NGN is widely expected to be supported by SDN to enhance the network utilization in terms of throughput. We propose a novel routing algorithm-CentFlow, for an SDN domain to boost up the network utilization. The proposed weight functions in CentFlow achieve smart traffic distribution by detecting highly utilized nodes depending on the centrality measures and the temporal node degree that changes based on node utilization. Furthermore, the frequently selected edges are penalized thereby augmenting the flow balancing and dispersion. CentFlow reaps greater benefits on an SDN controller than the classical OSPF due to its comprehensive view of the network. Experimental results show that CentFlow enhances the utilization of up to 62% of nodes and 49% of links, respectively, compared to an existing Dijkstra algorithm-based routing scheme in SDN. Furthermore, nearly 6.5% more flows are processed networ- wide.