Virtual Network Functions Placement and Chaining for real-time applications

Network Function Virtualization (NFV) has attracted a lot of attention recently both among researchers and network devices providers. By decoupling network functions from dedicated hardware devices and sharing physical network resources, NFV can allow a collection of multiple network services to run simultaneously in software modules over a shared network. However, a major challenge that needs to be addressed is the Placement and Chaining of the Virtual Network Functions (PC-VNF). This allows determining the optimal placement and the required instances of VNFs needed to deploy Service Function Chains (SFCs) while optimizing the cost of resources provisioning. In this context, real-time applications such as VoIP, network gaming, IPTV, and media streaming require a bounded end-to-end delay to meet quality of service requirements. It is therefore crucial that their needs be taken into consideration when planning for deploying SFCs. In this paper, we formalize the PC-VNF problem for real-time applications as an Integer Linear Programming (ILP) and we propose a heuristic solution in order to handle with scalability problem. Results show that the proposed heuristic has a better scalability comparing with the ILP model, with a reduction of up to 80% in execution time and provides an acceptable end-to-end delay suitable for real-time applications. Moreover, the average of accepted requests is still satisfactory depending on delay bound constraint.