Gerenciamento de tráfego dinâmico com ciência de energia para redes definidas por software

Abstract

Energy consumption is a central issue in computer networks nowadays. In most cases, networks resources are overprovisioned to handle overloaded traffic, which clearly incurs into waist of resources and high energy costs during the major time. However, it is possible to proper deactivate links and network devices, as switches, to reduce energy consumption during underutilization periods and, as consequence, decrease operational costs. In this work, we present an energy-aware dynamic traffic management mechanism for Software-Defined Networks (SDN). SDN allows a flexible network programmability through a centralized controller, which cannot be achieved by traditional architectures. The proposed mechanism explores this feature to optimize traffic routing throughout switches, disabling links and allowing the SDN controller to turn-off switches on the fly. Moreover, it dynamically tries to aggregate traffic flows, reducing the total number of endto- end paths, which allows a better network usage. We evaluated the proposed mechanism by emulating a realistic campus-based SDN environment. We considered different traffic levels, ranging from lightweight to overloaded scenarios. Results highlight the energy savings achieved by the proposed solution is 46.01% in a low load scenario, 36.72 % in a medium load scenario and 17.86 % in a considered overloaded scenario.