Controller Placement for Resilient Network State Synchronization in Multi-Controller SDN

Abstract

In a multi-controller SDN, the network state is replicated or distributed between the controllers requiring repeated synchronization. This synchronization cost depends on controller locations, as well as needs to be low and resilient to network failures. Leveraging a Steiner tree-based inter-controller latency model, a multi-objective integer linear program is presented to deduce the controller placement optimizing (a) synchronization cost in failure-free scenarios, and, (b) resilience against single-link failures. The proposed approach incurs low computational complexity and post-failure network reconfiguration cost, while achieving near-optimal post-failure synchronization cost. We extensively evaluate the proposed model over real networks, and demonstrate its effectiveness.