A Dual Decomposition Method for Hierarchical Traffic Control in Inter-DC WANs

Resource utilization of inter-data center wide area networks (inter-DC WANs) has been increased by software-defined networking (SDN) technology significantly, such as Google’s B4. However, the scalability of SDN network becomes more and more prominent with the increase of Cloud services. Google has improved network scalability through hierarchical network topology in the follow-up work of B4. In this paper, we focus on how to design an appropriate distributed algorithm for hierarchical traffic control in inter-DC WANs based on SDN. Compared with B4, which focused on satisfying stringent availability targets in every service class, this paper focuses on improving the utility of inter-DC WAN which is of wide sense for general traffic control by scheduling network background traffic based on theoretical modeling. We establish a model for the inter-DC WAN by considering the traffic priorities and formulate an optimization problem to maximize the utility function of the network. We propose a two-layer traffic control structure and design a distributed multi-domain control (DMC) algorithm by using dual decomposition method for hierarchical traffic control in inter-DC WANs. This distributed algorithm is implemented by exchanging the information between the local domains and the super controller. Simulation results of the case study with maximizing network throughput optimization objective verify the proposed algorithm by comparing with the centralized algorithm.