A Semantic virtualized network functions description and discovery model

International audienceNetwork Function Virtualization (NFV) has increasingly gained importance to address some emerging networking challenges like agility and cost-effectiveness. NFV enables to run Virtualized Network Functions (VNF) on top of any generic, Commercial-Off-The-Shelf (COTS) hardware, anytime and anywhere in the network. Specific service providers offer VNFs to prospective network providers. Service providers publish VNFs in dedicated marketplaces where network providers search VNFs and instantiate them according to a pre-established service-level agreement. On top of being proprietary and specific to the service providers, the existing VNF description models include details on VNF deployment but fail to fit VNF functional and non-functional specifications. This description alters an efficient selection of the most relevant VNFs and prevents full automation of the VNFs provisioning. This paper introduces a novel domain-independent VIrtualized networK functIoN ontol-oGy (VIKING for short) for VNF description and publication in federated repositories. It also proposes a semantic-based matchmaking algorithm to discover and select the most relevant VNFs that satisfy prospective VNF consumers' requests. As for validation, a prototype called Mastermyr Chest, including VIKING's instantiation along with the matchmaker in Content Delivery Networks (CDN) domain was implemented. This prototype illustrates a new way to contribute to the redesign of the CDN's traditional architecture by enabling value-added CDN service provisioning in an agile and dynamic manner. A set of experiments was run to (i) evaluate the matchmaker performances and (ii) demonstrate its accuracy and precision