The emergence of 5G enables a broad set of diversified and heterogeneous services with complex and potentially conflicting demands. For networks to be able to satisfy those needs, a flexible, adaptable, and programmable architecture based on network slicing is being proposed. Moreover, a softwarization and cloudification of the communications networks is required, where network functions (NFs) are being transformed from programs running on dedicated hardware platforms to programs running over a shared pool of computational and communication resources. This architectural framework allows the introduction of resource elasticity as a key means to make an efficient use of the computational resources of 5G systems, but adds challenges related to resource sharing and efficiency. In this paper, we propose Artificial Intelligence (AI) as a built-in architectural feature that allows the exploitation of the resource elasticity of a 5G network. Building on the work of the recently formed Experiential Network Intelligence (ENI) industry specification group of the European Telecommunications Standards Institute (ETSI) to embed an AI engine in the network, we describe a novel taxonomy for learning mechanisms that target exploiting the elasticity of the network as well as three different resource elastic use cases leveraging AI. This work describes the basis of a use case recently approved at ETSI ENI.

The future mobile networks have to be flexible and dynamic to address the exponentially increasing demand with the scarce available radio resources. Hence, 5G systems are going to be virtualised and implemented over cloud data-centres. While elastic computation resource management is a well-studied concept in IT domain, it is a relatively new topic in Telco-cloud environment. Studying the computational complexity of mobile networks is the first step toward enabling elastic and efficient computational resource management in telco environment. This paper presents a brief overview of the latency requirements of Radio Access Networks (RANs) and virtualisation techniques in addition to experimental results for a full virtual physical layer in a container-based virtual environment. The novelty of this paper is presenting a complexity study of virtual RAN through experimental results, in addition to presenting a model for estimating the processing time of each functional block. The measured processing times show that the computational complexity of PHY layer increases as the Modulation and Coding Scheme (MCS) index increases. The processes in uplink such as decoding take almost twice the time comparing to the related functions in the downlink. The proposed model for computational complexity is the missing link for joint radio resource and computational resource management. Using the presented complexity model, one can estimate the computational requirement for provisioning a virtual RAN as well as designing the elastic computational resource management.

The combination of Network Function Virtualisation (NFV) and cloud-based radio access network (C-RAN) is a candidate approach for the next generation of mobile networks. In this paper, the novel concept of virtual radio resources, which completes the virtual RAN paradigm, is proposed. The key idea is to aggregate (and manage) all the physical radio resources, to create virtual wireless links, and to offer Capacity-as-a-Service. Due to the isolation among instances, network element abstraction, and a multi-radio access techniques (RAT) structure, the virtualisation approach leads to relatively more efficient and flexible RANs than former ones. Virtual network operators (VNOs) ask for wireless connectivity in the form of capacity per service, hence, not dealing with physical radio resources at all. A model for the management of virtual radio resources is proposed, which can even support the shortage of resources. A practical heterogeneous cellular network is considered as a case study, and results are presented, showing how the virtual radio resource management allocates capacity to services of different VNOs, with different service-level agreements (SLAs) and priority when the overall network capacity reduces down to 45% of the initial one.