Simplifying Network Updates in SDN and NFV Networks Using GUM

As new network paradigms, software-defined networking (SDN) and network function virtualization (NFV) enable network innovation and brings flexibility in network management and service deployment. However, It is still challenging task to ensure that network policies remain consistent during the network updates due to the inherently distributed nature of the network. In this paper, we propose the Generalized Update Model (GUM) to support fast and consistent network updates under different levels of constraints, including connectivity consistency, policy consistency, and performance consistency. In our model, we use general high-level abstractions for capturing these consistent constraints and generate a state-resource dependency graph (SDG). With the help of the SDG, we analyze relations between the update operations and construct an operation relation graph (ORG) to find the optimal update operation sequence. We prototype GUM on Ryu and evaluate it by comprehensive emulations and data- driven simulations. The results show that our scheme can always obtain the best update operation sequence under different requirements and speed up the update process by 32% on average.