Optimal Virtual Network Function Deployment for 5G Network Slicing in a Hybrid Cloud Infrastructure

Network virtualization is a key enabler for 5G systems to support the expected use cases of vertical markets. In this context, we study the joint optimal deployment of Virtual Network Functions (VNFs) and allocation of computational resources in a hybrid cloud infrastructure by taking the requirements of the 5G services and the characteristics of the cloud architecture into consideration. The resulting mixed-integer problem is reformulated as an integer linear problem, which can be solved by using a standard solver. Our results underline the advantages of a hybrid infrastructure over a standard cloud radio access network consisting only of a central cloud, and show that the proposed mechanism to deploy VNF chains leads to high resource utilization efficiency and large gains in terms of the number of supported VNF chains. To deal with the computational complexity of optimizing a large number of clouds and VNF chains, we propose a simple low-complexity heuristic that attempts to find a feasible VNF deployment solution with a limited number of functional splits. Numerical results indicate that the performance of the proposed heuristic is close to the optimal one when the edge clouds are well dimensioned with respect to the computational requirements of the 5G services.

I. Introduction

The development of the fifth generation (5G) system is driven by the goal of addressing new mobile services characterized by heterogeneous requirements, e.g., peak data rate, latency, and network energy efficiency [2]. To fulfill this goal, the 5G research community is defining a new flexible architecture, where the network infrastructure is split into logical instances, i.e., network slices, each tailored to a dedicated service and running in a cloud infrastructure. A network slice could span across multiple domains, i.e., radio access network, transport network, and core network, and could be deployed across multiple operators. It comprises dedicated and/or shared resources and can be completely isolated from the other network slices to fulfill service level agreement [3].