Incremental Hybrid SDN Deployment for Enterprise Networks

Abstract

Software-defined networking (SDN) is a centrally managed network architecture which can flexibly allocates network resources according to current network conditions. However, since the cost of replacing all legacy switches to commercial SDN devices is not affordable to most enterprises, only a few switches can be replaced at a time. Such network architecture, combining legacy and SDN switches, is named as hybrid SDN. In this work, we formulate the hybrid SDN deployment problem as an optimization problem with resource constrains. Since this problem is NP-hard, we relax this problem to allow the use of greedy-based heuristic algorithms and propose a switch deployment strategy named HybridScore. HybridScore aims to provide the most control capacity over the network given a fixed number of switches to upgrade in hybrid SDN environment. The proposed HybridScore exploits the real traffic pattern of the network to determine the location of new SDN switches. The simulation results conducted with real-world enterprise network topology demonstrates that compared with existing strategies, our HybridScore deployment strategy improves the result by 11.28%. Moreover, compared with full SDN environment, our HybridScore provides comparable controlling capacity with only about 13% SDN switches.