No abstract

Network slicing is one of the key enabling techniques for 5G, allowing Network Providers (NPs) to support services with diverging requirements on top of their physical infrastructure. In this paper, we address the limited support and oversimplified resource allocation on different network segments of existing End-to-End (E2E) orchestration solutions. We propose a hierarchical orchestration scheme for E2E networks, breaking down the E2E resource management and network slicing problems per network segment. We introduce a higherlevel orchestrator, the hyperstrator, to coordinate the distributed orchestrators and deploy Network Slices (NSs) across multiple network segments. We developed a prototype implementation of the hyperstrator and validated our hierarchical orchestration concept with two proof-of-concept experiments, showing the NS deployment and the impact of the resource allocation per network segment on the performance of NSs. The results show that the distributed nature of our orchestration architecture introduces negligible overhead for provisioning NSs in our particular setting, and confirm the need of a hyperstrator for coordinating network segments and ensuring consistent QoS for NSs.

No abstract

O advento de redes definidas por software e virtualização de funções de rede trouxe inúmeras vantagens; contudo, para atingir a flexibilidade e a programabilidade previstas nessas tecnologias, novos componentes nos planos de controle e gerenciamento foram introduzidos. Tais componentes requerem recuperação rápida pois, sem gerenciamento, todo o plano de dados fica inoperável. Para lidar com falhas nesses planos, recorre-se à técnica de autocura, explorada na tese que é resumida neste documento. Os resultados provam a eficácia de autocura em fatias de rede com rigorosos requisitos de qualidade e, também, demonstram que o framework introduzido é capaz de autocurar, ou seja, curar o ambiente degradado bem como curar a si mesmo.