Service Chain Embedding for Diversified 5G Slices With Virtual Network Function Sharing

We address the problem of embedding service chains consisting of a sequence of virtual network functions (VNF) for 5G slices, considering diversified slice requirements. We develop a fine-grained approach that considers resource requirements and limited traffic processing capacity of VNFs, which can be shared (or not) among slices depending on VNF functionalities. We develop an optimization formulation that minimizes the resources while satisfying the constraints in both the physical infrastructure and logical slices. We also develop a heuristic algorithm based on a goodness function to overcome the problem complexity. The numerical results show that our approach uses 30% less bandwidth and 45% fewer VNF instances, compared with a baseline.