Modelo para garantir o atraso máximo para conexões VoIP em uma rede SDN ODL

Abstract

Traditional networks work in a distributed manner, where each element has a control plane. They have disadvantages such as: decision-making is distributed across several network components and equipment from different manufacturers each has its own configuration interface, increasing the complexity and time of equipment configuration. To solve these problems, SDN networks emerged, where the data plane, responsible for forwarding data, is separated from the control plane, responsible for configuring and managing the network elements, being represented by a controller. For communication between the controller and network components, the OpenFlow protocol is used. As with traditional networks, SDN networks need to provide QoS to users, which is a mechanism to improve performance in a communication network and manage network resources more efficiently. An SDN network can provide QoS through two tools that have been implemented until the moment of the development of this work: meter and queue. Meter is a mechanism used at the network entrance to limit the data rate and queue is used at the network exit where multiple queues can be configured to separate the traffic. As there is still no defined model using these two mechanisms to provide QoS in a network with different traffics, in this work a model was developed to determine the queue size as a function of the number of users and the maximum delay for VoIP connections over an SDN network. The proposal was developed through simple and multiple linear regression and the results with the developed equation were compared with simulations in the mininet simulation environment with ODL controller. The results show that the queue size is directly proportional to the number of VoIP connections and the theoretical values are close to the values obtained through the simulation.