CUSCO: A Customizable Solution for NFV Composition

Abstract

Although Network Function Virtualization (NFV) has multiple advantages in comparison with traditional hardware middleboxes, there are still many open problems. Some of the major challenges are related to the service deployment process (composition, embedding, and scheduling). In particular, current solutions for network service composition are limited, in the sense that they are not customizable, neither in terms of the evaluation setup nor the operational behavior. In this paper, we propose a new adaptive service composition solution that takes into account multiple specific requirements of network operators. The proposed solution uses a statistical method to conciliate different metrics, disparate granularities, and conflicting objectives, and returns a composition result that maximizes the cost-benefit. We present a case study and experiments to show the feasibility of the proposed solution.