A network operator-biased approach for multi-service network function placement in a 5G network slicing architecture

The 5G communication standard is characterized by an increased softwarization, allowing a higher flexibility able to cope with different requirements and services. In particular, Network Function Virtualization (NFV) is a recently introduced technology that enables a software implementation of different network functions exploiting virtualization techniques, hence, enabling their flexible deployment upon system requirements. Boosted by NFV, the concept of network slicing is gaining great attention in 5G networks. The idea is that physical communication and computing resources are sliced in multiple end-to-end logical networks, each one tailored to best support a specific service. The advantages of NFV, in the network slicing context, are even more evident in distributed computing environments, such as the edge-to-cloud continuum, recently introduced for enabling a flexible deployment of multiple functions. In particular, thanks to the introduction of cloud-native technologies, based on the usage of containerization and microservice technologies, the virtual network functions (VNFs) deployment and their orchestration is an easy operation, allowing the on-the-fly network configuration. Gaining from the NFV, Network Slicing and Edge-to-Cloud continuum paradigms, we propose a new network function allocation problem for multi-service 5G networks, able to deploy network functions on a distributed computing environment depending on the service requests. The proposed approach jointly considers Radio Access Network (RAN) and Core Network (CN) functions and, dierently from other approaches, introduces an option able to bias the function placement depending on the service requirements, allowing a fast-and-easy operator-side deployment of the network functions. We propose to solve the problem through a Genetic Algorithm able to approach the optimal solution but with reduced complexity and execution time. The performance is compared with two other heuristic algorithms and with an exhaustive search algorithm, introduced as benchmarks, showing the benefits of the selected solution in terms of performance, flexibility and complexity.