Network slicing represents a substantial technological advance in 5G mobile network, greatly expanding the variety and manifoldness of network services to be supported. Additionally, 3GPP 5G New Radio (NR) has introduced novel features such as mixed numerology and mini-slots, which can be harnessed by network slicing to cater to the diverse requirements of 5G services. While however the co-existence of multiple network slices leads to a challenging resource allocation problem, these new features also severely complicate the management of radio resources. As a further point of attention, the virtualization of radio functions may exact a significant toll from the, already limited, computing resources at the network edge. It follows that a cost-efficient resource allocation across all the slices becomes crucial. In this paper, we address the above-mentioned issues by modeling a cost-efficient radio resource management in 5G NR featuring network slicing, named CERS, through a Mixed Integer Quadratically constrained Program (MIQCP). We maximize the profit of all slices simultaneously guaranteeing the target data rate and delay specified in the service level agreements (SLAs) fo the different traffic flows. To reduce the complexity of the MIQCP problem, we decompose it into two sub-problems, namely, the scheduling problem of enhanced Mobile Broadband (eMBB) user equipments (UEs) on a time-slot basis and of Ultra-Reliable Low Latency Communications (uRLLC) UEs on a mini-slot basis, while keeping the objective unchanged. To address the scheduling issue of eMBB UEs, we employ a heuristic technique, and, by leveraging the outcome of this heuristic, we derive an optimal solution for the problem of uRLLC UEs. The significance of the proposed approach over a baseline approach is evaluated through extensive numerical simulations in terms of the number of allocated uRLLC resource blocks (RBs) per mini-slot. We also assess our approach by measuring the impact of the uRLLC slice changes on the eMBB slice, and vice versa, including delay for uRLLC users and data rates for eMBB users.

The combination of service virtualization and edge computing allows for low latency services, while keeping data storage and processing local. However, given the limited resources available at the edge, a conflict in resource usage arises when both virtualized user applications and network functions need to be supported. Further, the concurrent resource request by user applications and network functions is often entangled, since the data generated by the former has to be transferred by the latter, and vice versa. In this paper, we first show through experimental tests the correlation between a video-based application and a vRAN. Then, owing to the complex involved dynamics, we develop a scalable reinforcement learning framework for resource orchestration at the edge, which leverages a Pareto analysis for provable fair and efficient decisions. We validate our framework, named VERA, through a real-time proof-of-concept implementation, which we also use to obtain datasets reporting real-world operational conditions and performance. Using such experimental datasets, we demonstrate that VERA meets the KPI targets for over$96\%$of the observation period and performs similarly when executed in our real-time implementation, with KPI differences below 12.4%. Further, its scaling cost is$54\%$lower than a centralized framework based on deep-Q networks.

Network slicing is a promising technique that has vastly increased the manifoldness of network services to be supported through isolated slices in a shared radio access network (RAN). Due to resource isolation, effective resource allocation for coexisting multiple network slices is essential to maximize network resource efficiency. However, the increased network flexibility and programmability offered by virtualized radio access networks (vRANs) come at the expense of a higher consumption of computing resources at the network edge. Additionally, the relationship between resource efficiency and computing cost minimization is still fuzzy. In this paper, we first perform extensive experiments using the vRAN testbed we developed and assess the vRAN resource consumption under different settings and a varying number of users. Then, leveraging our experimental findings, we formulate the problem of cost-efficient network slice dimensioning, named cost-efficient slicing (CES), which maximizes the difference between total utility and CPU cost of network slices. Numerical results confirm that our solution leads to a cost-efficient resource slicing, while also accomplishing performance isolation and guaranteeing the target data rate and delay specified in the service level agreements.

The combination of service virtualization and edge computing allows mobile users to enjoy low latency services, while keeping data storage and processing local. However, the network edge has limited resource availability, and when both virtualized user applications and network functions need to be supported concurrently, a natural conflict in resource usage arises. In this paper, we focus on computing and radio resources and develop a framework for resource orchestration at the edge that leverages a model-free reinforcement learning approach and a Pareto analysis, which is proved to make fair and efficient decisions. Through our testbed, we demonstrate the effectiveness of our solution in resource-limited scenarios, and show an improvement of around 60% in the CPU budget violation rate with respect to RL based standard multi-agent framework.