Stability of virtual network topology control for overlay routing services

Abstract

Overlay networks achieve new functionality and enhance network performance by enabling control of routing at the application layer. However, this approach results in degradations of underlying networks due to the selfish behavior of overlay networks. We discuss the stability of virtual network topology (VNT) control under overlay networks that perform dynamic routing updates. We find that the dynamics of routing on overlay networks cause a high fluctuation in the traffic demand matrix, which leads to significant VNT control instability. To overcome this instability, we introduce three extensions, hysteresis, two-state utilization hysteresis, and filtering, to VNT control. Simulation results show that the hysteresis mechanism improves network stability, but cannot always improve network performance. We therefore extend the hysteresis mechanism and show that it improves both network stability and performance. However, this extension requires a lot of time for the VNT to converge to a stable state. To achieve fast convergence, we use a filtering method for VNT control. Through simulations, we prove that our methods achieve stability against overlay routing without loss of adaptability for changes in traffic demand.