Toward the Optimal Service Function Chain Placement in a Multi-Domain 5G Architecture

5G Connected Automated Mobility (CAM) offers broad services (automotive driving, V2X applications, infotainment etc.). Cross border scenarios that ensure uninterrupted CAM services in multi-domain environments is of paramount importance. The 5G communication networks form the basis for the deployment of CAM services, aiming to provide guarantees across borders. Accordingly, we propose a multi-domain core-edge tier 5G architecture able to support 5G CAM services, also supporting migrations. The services are deployed as Service Function Chains (SFCs), namely a set of interconnected Virtual Network Functions (VNFs). Different VNFs of the same SFC can be placed in the same or different domains thanks to cross domain Core-to-Core and MEC-to-MEC (Multi-Access Edge Computing) communication. The VNF placement decision is based on latency and cost constraints, specific per VNF, which determine if a VNF migration is allowed. Moreover, latencies between any two connected hosts, that accommodate a VNF or a planned migration, are taken into consideration. It is highlighted that the multi-domain MEC-to-MEC communication along with the consideration of VNFs service migrations is an emerging approach. In order to test the functionality of our proposed architecture, we formulate the problem as an Integer Linear Programming (ILP) optimization problem which aims to the minimization of the aforementioned latency and cost constraints. Moreover, we also propose an innovative VNF placement heuristic algorithm that is compared against the ILP model, assuming an “online” version of the problem: requests arrive in time, and the total number of the SFC placement requests is not known in advance. Our experimental results showcase that our heuristic algorithm runs efficiently while approximates well the optimal. The ILP constitutes the most cost-effective solution, while the heuristic runs on lower complexity and execution time with slightly higher deployment cost.