Real-time Dynamic Network Slicing for the 5G Radio Access Network Massimiliano Maule*, Prodromos-Vasileios Mekikis*, Kostas Ramantas*, John Vardakas*, and Christos Verikoukis† *Iquadrat Informatica S.L., Barcelona, Spain †Telecommunications Technological Center of Catalonia (CTTC/CERCA), Castelldefels, Spain Abstract—The 5G networks are expected to satisfy diverse use cases and business models with significant advancements in terms of capacity, reliability, and latency. The allocation and provisioning of network resources pose a challenge for this novel architecture to guarantee higher flexibility and quality of service. As a potential enabler, network slicing was proposed as an innovative approach for the control of the network resources. Although a static slicing approach can be suitable for the transport and core network, the stochastic behavior of the wireless channel requires fast and secure slicing techniques for resource allocation. In this paper, we propose a dynamic slicing approach for the radio access network, where the network resources are carefully assigned to guarantee the service level agreements and increase the number of served users. To prove the performance of our approach, we implemented a fronthaul testbed to emphasize the strength of our method in terms of throughput and resource utilization, compared to static slicing. Index Terms—5G network, RAN slicing, Software-defined Network, Software-defined Radio, LTE, virtualization I. I NTRODUCTION Expected to be deployed first in dense urban areas, 5G networks will cover over 20 percent of the world’s population by the end of 2023 [1]. Mobile data traffic is expected to surge by eight times during the forecast period, reaching 110 exabytes per month by 2023 [1]. To achieve an ultra-high data transmission rate and an ultra-low response times (latency), 5G will be deployed through technologies, as Software De- fined Networking (SDN) and Network Function Virtualization (NFV). SDN and NFV represent the key technologies for efficient 5G network management. With SDN, a network is capable to support the traffic management needs dictated by the new forms of distributed processing. Moreover, NFV introduces the concept of virtualization of Network Functions (NFs), translating hardware-based appliances into high volume servers, switches and storage. A significant aspect of this new architecture concerns the provision of a wide range of services with different require- ments. The control of this architecture is based on three classes of services (i.e., enhanced mobile broadband (eMBB), ultra-reliable and low-latency communication (URLLC) and massive machine-type communications (mMTC)) in which the flows are guided according to their requirements [2]. To perform these type of services on the new 5G architecture through SDN and NFV techniques, the concept of Network Slicing (NS) has been proposed. The term network slice refers to a network with specific resources and functions, which are perceived by the user as if they were a dedicated physical network, isolated from other virtual systems [3]. A service- oriented slice approach represents the perfect technique for isolating specific functionalities, guaranteeing a certain type of resources and maintaining the continuity of the end-to-end service on top of the Physical Network (PN). To that end, there are two ways to perform slicing within a network. Static slicing consists in assigning a fixed and constant portion of physical resources to the virtual network for the duration of the service. Consequently, this approach can lead to a waste of resources, as the load in the network is variable and influenced by different factors (e.g., difference in use between day and night, flux variations by geographical area). On the other hand, dynamic slicing consists in assigning resources to a specific virtual network based on the type of service requested. In the current state of the art, NS is mainly focused on a static approach since the current network architecture model does not fully support NFV and SDN techniques necessary for the correct establishment of the slices. In [4], a prototype of an end-to-end (E2E) NS testbed is proposed, where the importance of correct slice parameterization is discussed. Moreover, the work in [4] illustrates the limitations of static slicing and the need to introduce dynamic techniques. Different models based on machine learning and artificial intelligence have been proposed for the derivation of the op- timal dynamic slicing approach. In [5], the authors formulate the dynamic slicing problem as a Mixed Integer Linear Pro- gramming (MILP) problem. Also, in [6], authors focus on the Deep Q-Learning technique for slicing resource management while in [7] the authors propose a new dynamic NS scheme based on Baseband Unit (BBU) capacity allocation and Phys- ical Resource Blocks (PRBs) management. According to our analysis, even though these approaches provide the optimal solutions for dynamic slice selection and configuration, they are far from a physical application in a real scenario, where the communication channel is subjected to sudden variations that can significantly compromise learning models based on the network evolution. In this paper, we present a dynamic slicing approach whose objective is to minimize the waste of resources by guaranteeing the minimum requirements of each service requested by the users. The innovative part of our solution is the rapid slice parameterization changes according to the traffic requirements and the management of unpredictable variation due to the