Summary: | The Internet is currently essential in all its functions to every social layer. The fast and unexpected growth of the largest computer network combined with the gradual change of its main objective forced service providers to nd new solutions to improve and optimize their networks and equipments. The craving for bigger and better services by the users increased the use of Cloud services, demanding computing and network automation, notably in datacenters. After the development of virtualization and the popularity rise in Cloud Computing (CC), the dynamization of the datacenters began. However, although the new paradigm increased the response time and scalability of cloud service provider's infrastructure, the network became the performance bottleneck. To solve this shortcoming, Software De- ned Networks (SDN) appeared. This is a new approach to the classic computer networking paradigm that decouples the control plane from the data plane. Although it is a new technology, its development and usage is having a great acceptance by the market and by big telecommunication and cloud service provider players. At the same time, the development of network virtualization enabled network element abstraction from the hardware. This change granted exibility and scalability in network management that was not available before. With these evolutions in mind, this dissertation objective is to study, develop and analyse an example of a provider core network using Open- ow technology to establish cloud service connections. The rst objective is the study of the existing SDN solutions and the analysis of its capabilities. After this step, a SDN module has been developed to control the network. This application allows users to establish a service in a simple and safe way in the provider network, with easy automation for management. Afterwards, a platform has been deployed and test scenarios have been performed to determine the performance of the implementation in a provider network. To analyze the performance in large-scale networks, scalability tests have been performed in emulated environments.
|