| dc.description.abstract | Network Function Virtualization has been brought up to allow the TSPs to have more possibilities and flexibilities to provision services with better load optimizing, energy utilizing and dynamic scaling. Network functions will be decoupled from the underlying dedicated hardware into software instances that run on commercial off-the-shelf servers. However, the development is still at an early stage and the dependability concerns raise by the virtualization of the network functions are touched on only briefly. Particularly, the evaluation of the NFV-based services dependability has never been conducted.
Towards this goal, this thesis aims to address the dependability concern of NFV and uses a two-level availability approach to construct a quantitative evaluation about how to assess the availability of an NFV-based service and how NFV elements in the network shall be deployed to provide a more dependable network service.
A two-level availability model has been developed based on the various NFV-based network. The first level is focusing on the topology of the network and the connectivity requirements for provisioning an NFV-based service. In the second level, the Stochastic Activity Network (SAN) model of different network elements such as VNF, NFV-MANO and datacenter have been developed to evaluate the availability. Eventually, these two types of models have been merged together to illustrate the overall availability/unavailability of the NFV-based services in different use cases.
In the end, analysis and evaluation have been conducted based on the obtained results from the two-level availability model. There are seven different scenarios have been simulated with regard to the deployment of NFV across the network. And the outcome on how the variations of the NFV elements deployment influence the dependability of the NFV-based services will be presented along with some suggestions about the NFV deployment in provisioning an end-to-end service. | |
