Modeling and Provisioning Highly Available NFV Services

Abstract

The ubiquitous deployment of middleboxes can hamper the network capability to be flexible, scalable and innovative to an extent that new and specialized services cannot be easily introduced in the network. Network Function Virtualization (NFV) promises to overcome these limitations by providing the ability to execute virtual instances of networking functions on top of a common physical network substrate. It moves data processing tasks from proprietary hardware middleboxes to virtualized entities that can run on commodity hardware. Together with Service Function Chaining, it enables the replacement of traditional network hardware appliances with softwarized Virtualized Network Function (VNF) chains. However, this major transformation brings new challenges and a crucial one is the ability to ensure the high-availability demands of carrier-grade services provided by NFV-enabled networks. This challenge is further exacerbated by the extreme availability levels that 5G use cases demand, e.g., ultra-reliable services. This work tackles the challenge by addressing the problem of how to assess and quantify the availability of NFV-supported services, and how to provision highly available services by means of fault-tolerant mechanisms that are both effective and resource efficient.