5G and beyond technologies are designed to convey multiple QoS requirements coming from highly heterogeneous services. In this regard, 5G extends the QoS definition of previous technologies. In order to fulfill the requirements of new services, it is necessary to develop resource management solutions that effectively and efficiently translate them to usage of resources. Therefore, MAC schedulers are to be designed to accommodate the demands of emerging applications with stringent QoS requirements, such as XR. In this work, we introduce a Lyapunov-based MAC scheduler designed to appropriately tackle coexisting heterogeneous QoS flows. The proposed approach allows the implementation of policies that explicitly consider the QoS requirements, as well as an efficient use of allocated resources. The results obtained through extensive experiments over the ns-3 5G-LENA simulator demonstrate that our scheduler guarantees the required time-average bit rate for every QoS flow at every time window, while ensuring the stability of traffic queues for all users.

XR applications place increasing pressure on mobile networks to meet their stringent Quality of Service (QoS) demands. In particular, XR services require high throughput and low latency, with traffic patterns that differ greatly from traditional applications. To address these challenges, we propose a novel Medium Access Control (MAC) scheduling solution that takes into account the segmentation of XR media units into Packet Data Units (PDUs), referred to as PDU sets in 5G-Advanced networks. These PDU sets correspond to the core traffic unit for XR services and must be delivered according to their specific QoS requirements, which are dynamically established in real-time by the XR application. The proposed scheduling solution leverages Lyapunov drift-plus-penalty optimization to jointly optimize resource allocation and stabilize traffic queues while ensuring the timely delivery of PDU sets. Our solution is implemented within the ns-3 5G-LENA framework, and the paper features a comparative analysis through extensive simulations, evaluating the performance of our proposal against existing MAC scheduling algorithms. Results evince that the proposed scheduler enhances resource utilization, ensures QoS compliance, and improves network performance.

In 5th Generation (5G) virtualized Radio Access Networks (vRAN), where the protocol stack is split across three different elements (Radio Unit (RU), Distributed Unit (DU), and Centralized Unit (CU)), and data is transmitted across the transport network (fronthaul (FH) and midhaul) between these units, the capacity of the fronthaul pipe can become a limiting factor for packet transmission. To address this challenge, state-of-the-art approaches have introduced several fronthaul control methods, designed to limit the sending rate, to ensure that it does not exceed the available capacity. However, it is important to avoid potential conflicts between the decisions of such fronthaul control methods and those of the Medium Access Control (MAC) scheduling schemes, as they might impact one another. In this context, we provide a thorough analysis of the trade-off between the fronthaul and the air interface limitations and we propose a new FH Control method to limit packet transmissions without altering the decisions made by the MAC scheduler. An extensive simulation campaign in a 3GPP defined scenario with Virtual Reality (VR) and Cloud Gaming (CG) services is conducted, for air-limited and air/fronthaul capacity-limited conditions. The results demonstrate that the proposed fronthaul control method efficiently coordinates the MAC scheduler decisions with the constraints imposed by the fronthaul capacity.

We jointly consider network slicing and functional split for 5G/6G Radio Access Network (RAN). Exploiting virtu-alization techniques, virtual Radio Access Network (vRAN) is an architectural shift that moves some of the functions that were traditionally performed by the base station to centralized nodes, deployed at cloud-based resources. By appropriately selecting the location of these nodes, they can effectively coordinate a number of radio access elements, thus bringing potential gains, for instance, in terms of interference reduction. One pivotal decision for vRAN operation is how to split the functions between these central elements and the corresponding distributed units. Deploying more functions at the cloud could potentially bring more benefits, due to the tighter cooperation between access elements. On the other hand, higher centralization poses more requirements, in terms of computational resources at the nodes where these functions are deployed. Along with the virtualization, RAN slicing enables the deployment of independent service verticals over the same physical infrastructure, and this could greatly benefit from the coordination achieved by exploiting the vRAN features. In this paper we propose a vRAN model to jointly consider functional split and RAN slicing, and we establish the corresponding configuration, exploiting Lyapunov's Theory to tackle the underlying problem. The results show that, in highly heterogeneous networks, the dynamic configuration of the functional split can reach the same performance of fully centralized networks, with substantially fewer computation resources.