Intelligent Control Plane Design for Virtual Software-Defined Networks

In virtual Software-Defined Networks (vSDN) - where multiple tenants share the same physical SDN topology - control plane resilience has utmost importance. Although previous network hypervisor placement methods were able to provide single link and hypervisor failure resilience with short control paths, the placement had to be frequently adjusted if the vSDN requests changed dynamically. In this paper we propose two approaches which follow an intelligent design principle, i.e., besides maximizing acceptance ratio of current vSDN requests they also prepare the placement for the future. Our first method uses a representative request set and the network flexibility metric to maximize preparedness to unknown failures and traffic changes. Our second approach proposes a self-adjusting control plane, where for each new request set a single hypervisor instance and its corresponding control path might be migrated to a new location. We conduct thorough simulations to investigate the performance of the approaches, and demonstrate that they significantly outperform previous methods from the literature.