ROND: Rethinking Overlay Network Design with Underlay Network Awareness

Abstract

Overlay network is an effective way for constructing end-to-end high-quality network services. Service Overlay Network (SON) leverages existing data transmission infrastructure to offer overlay network as network services, which provides flexibility and controllability for deploying value-added services. In this paper, we study the overlay topology design problem of SON deployment, that is, purchasing bandwidth with certain Quality of Service (QoS) guarantees in the underlay network, and building a logical overlay network that adapts to communication requirements with guaranteed reliability. The challenge lies in the difficulty of designing an overlay network with adaptability for dynamic of traffic requirement and uncertainty of underlay network failures. Traditional methods lack the underlay network view and rarely consider network failure scenarios, making it difficult to achieve global optimization. We mathematically formulate the minimum-cost overlay network design problem with network reliability constraints. Then, we propose a novel method, ROND, that samples critical traffic matrices to capture the dynamic of traffic patterns and generates failure scenarios to characterize the volatility of underlay network for avoiding significant network reconfiguration. On this basis, we simplify overlay network design problem and solve it iteratively, thereby minimizing the transmission cost as much as possible while ensuring the network performance. Extensive experimental results demonstrate that ROND is cost-effective and performance-guaranteed.