Autocura para redes definidas por software

Abstract

Software-defined Networking (SDN) and Network Functions Virtualisation (NFV) approaches have been improved in the last years, and both have advantages, mainly because of the logically centralised control plane view. SDN introduces some new components, e.g. SDN controller and SDN applications, while in traditional networks, these elements did not exist. Some works point problems related to failures in SDN and its components, e.g. forward rules inconsistency and software or hardware failures. Concerning NFV, the efforts to virtualise classic network functions bring a considerable complexity to manage the architectural components, e.g. resource orchestration systems. In both approaches, there is no architectural structure prepared to deal with management. The network management field has been structured after the creation of the approaches. As manual network management in these complex architectures is not suitable, we suggest a new management architecture: Self-organising Networks Architecture (SONAr). Such new architecture bases on autonomic computing fundamentals from software engineer and Self-organising Networks (SON) concepts from telecommunications for employing network self-management. In this thesis, we present the SONAr and extend the self-healing fundamental for SDN and NFV by creating a SONAr-based framework, i.e. Network Operations Self-healing (NOSH). We deeply investigated the self-healing fundamental considering SDN and NFV particularities, including the fault tolerance peculiarities from Data, Control and Management layers. We developed two distinct implementations, and the evaluation proves that NOSH heals SDN and NFV layers from performance degradations and heals NOSH itself. Additionally, the evaluation demonstrates improvements in Network Slices (NSs) with strict Quality of Service (QoS) requirements, e.g. in Ultra-Reliable and Low Latency Communications (URLLC), at run-time by using self-healing capabilities.