Abstract

In recent years, Network Function Virtualization (NFV) has become very popular for many reasons including flexibility and cost efficiency. Virtual Network Function (VNF) is the implementation of network functions using software which is decoupled from the underlying hardware. Content Delivery Network serves content to end-users with high availability and high performance. In addition to content, they also offer value-added services (video overlay, compressed video, video transcoding, etc.) to end-users. Publication and discovery mechanism is very much substantial to provide meaningful services to the end-users. Incorporating publication and discovery mechanisms in NFV can play a crucial role in CDN to provide on-demand value-added services. However, publication and discovery mechanisms for VNF remains a big challenge. As the existing researchers do not consider an exhaustive way of describing Virtual Network Functions (VNFs), VNF descriptors include details concerning VNF deployment but fail to include functional and non-functional specifications of VNF, which increases an inefficient utilization of VNFs. Another challenge is that the current discovery requires VNF consumers to select the VNF manually rather than having an automated mechanism to discover VNF. Furthermore, publication of VNF and its details are proprietary, which brings difficulty for VNF consumers in discovering and utilizing them. This thesis aims to propose a solution for efficient description, publication, and discovery of VNFs.
The main contribution of this thesis is twofold. First, we have proposed a VNF descriptor model including its properties and operations for efficient utilization. Second, we have proposed an architecture for publication of VNF descriptors and for discovering them. The proposed architecture use Domain Name System as a foundation. We have built a proof-of-concept prototype for value added service provisioning of VNF using the BIND open source server. In order to evaluate the viability of the proposed architecture, we have made performance measurement of the implemented prototype.