Combining Traffic and Link Information to Deploy SDN Switches in Hybrid SDN Enterprise Networks

Software-Defined Networking (SDN) is an emerging architecture that switches are centralizedly controlled. Unlike legacy architecture, SDN''s direct programmability brings a lot of benefits, such as easier management and more complicated network applications. However, changing existing network into SDN is not an ""one-click"" feature. The budget constraint is the primary limitation, especially for introducing SDN into enterprise networks. Therefore, we may instead use hybrid SDN, an architecture combining legacy and SDN switches. We find that there are two important challenges in hybrid SDN: how to measure traffics and where to deploy SDN switches. In this paper, we try to modify an existing network measurement for pure SDN network to be applied in hybrid SDN enterprise network, which just shows the SDN benefits in hybrid SDN network and records traffic traces for reference of future deployment. Then, we presents a greedy-based deployment strategy, which combines the information of traffic volume and node degree for switch selection. We demonstrate the feasibility of our modified network measurement by emulation. Our results suggest that we can get real-time measurement with low error rate. On the other hand, we evaluate our deployment strategy by simulation on real-world enterprise topology, and it shows that with about 25% aggregation/distribution switches upgraded to SDN, we can make whole network operate like it is a pure SDN. If we further limit the selection of routing path, we can achieve this goal by upgrading even less than 15% switches.