On-demand, dynamic and at-the-edge VNF deployment model application to Web Real-Time Communications

In the context of Network-as-a-Service (NaaS) for network operators openness towards Over-The-Top (OTTs) players, we propose an “On-demand and Dynamic Deployment Model for Virtual Network Functions (VNFs)”. We consider both, Telco's needs for more dynamicity in network service delivery, and OTTs' needs for network functions that support requirements of applications at the network-level. In order to allow third-party applications to request deployment of VNFs for specific needs, the proposed deployment model is based on an exposition approach that uses Network APIs over Network Functions Virtualization (NFV) Management and Orchestration modules. To illustrate the proposed model, we focus on Web Real-Time Communications (WebRTC). WebRTC applications are versatile and have strong needs for network functions which makes them a relevant use-case. We have implemented the proposed approach on OPNFV platform. We exploit the “at-the-edge” feature and location-awareness to respect the time-sensitivity and QoS constraints of WebRTC communications. For faster VNF instantiation, we rely on container-based virtualization (Docker containers). Finally, we perform experiments and evaluate the advantages of this model comparing to legacy and alternative approaches.