Network Virtualization (NV) has proved a promising technology that allows multiple heterogeneous Virtual Networks (VNs) to operate simultaneously on the same infrastructure. Dynamic Virtual Network Embedding (NVE) has emerged as an enabler of elasticity and scalability in the VN deployment and resource allocation of the physical infrastructure. However, the key challenge in realizing NV in a sustainable way is how to dynamically embed VNs efficiently into physical network, which is defined as the VNE problem. To address this challenge, this paper proposes an approach that leverages Multi-Agent Reinforcement Learning (MARL) to solve the dynamic VNE of VNs for 5G/6G communication systems. The proposed framework consists of multiple horizontally distributed RL agents that co-operate to devise temporally dynamic VNE placements. The key contributions of this work are introducing a novel dynamic VNE orchestration framework for multi-domain networks based on Distributed RL, providing a scalable VNE framework targeted to Ultra-Reliable Low-Latency Communication (URLLC) services, evaluating and comparing the proposed algorithm with existing solutions in the state of the art. The paper concludes that there is a significant improvement in latency of 144.191% when compared to the baselines.

As commercial 5G networks become commercially available and 6G looms in the horizon, the adoption of this new technologies depends on the way current and yet-to-come vertical industries put them to use. This accelerates the process of adoption by the end users, which are the final consumers of these technologies. 5G networks have multiple use cases associated with its new features, being the Ultra-Reliable Low Latency Communications (URLLC) use case one of the most instrumental for verticals such as remote teleoperation, factory automation and autonomous driving. In this paper, we design a general purpose and low-cost end-to-end (E2E) 5G/Beyond-5G Experimental Platform for URLLC applications, supporting Platform-as-a-Service (PaaS) with Artificial Intelligence (AI) and Machine Learning (ML) capabilities for online data analytics and automated decision making. The experimental evaluation of our platform demonstrates an average one-way latency on the DL and UL as low as 4.1 ms and 6.60 ms, respectively, 13.8 ms of end-to-end latency (E2EL), and a 31.7 ms E2EL between UEs for data frames of a teleoperation Web App with video feedback, demonstrating the capability of our platform in relation to other state-of-the-art testbeds for URLLC applications.

Managing network slices in 5G networks and in communication technologies Beyond-5G (B5G) requires intelligent mechanisms to ensure users’ service access and to maximize the utility and efficiency of the network’s physical resources. To achieve this, we propose a mechanism based on Reinforcement Learning (RL) for the Admission Control (AC) of User Service Requests (USRs) into network slices through dynamic bandwidth (BW) reallocation. Our approach admits, delays or rejects USRs into service depending on the BW of the slice and the utility this generates for the infrastructure provider (InP). This approach achieves very low USR rejection rates (RRs) with very high resource efficiency, even when peak traffic loads considerably exceed the BW capacity causing resource scarcity scenarios. When compared against a static BW allocation mechanism, our approach achieves RRs that are a fraction (0,33) of those achieved by static (smaller RRs are better), with 33,2x less resource overallocation, significantly achieving a very high resource efficiency.

The evolution of Software-Defined Networking (SDN) and Network Function Virtualization (NFV) in the telecommunications industry have intensified the issues of network management at large scales. Dynamic service orchestration and adaptive resource allocation became a necessity for network operators to manage the rapid growth of users and data-intensive applications. The impact of network automation on energy consumption and overall operating costs is often overlooked. Guaranteeing strict performance constraints of Ultra-Reliable Low Latency Communication (URLLC) services while enhancing energy efficiency is one of the major critical problems of future communication networks, given the urgency to reduce carbon emissions and energy consumption. In this work, we study the problem of zero-touch Service Function Chain (SFC) orchestration for multi-domain networks, targeting the latency reduction of URLLC services while improving energy efficiency for beyond-5G networks. Specifically, we propose SCHE2MA, a Service CHain Energy-Efficient Management framework based on distributed Reinforcement Learning (RL), that can intelligently deploy SFCs with shared VNFs per se into a multi-domain network. Finally, we evaluate SCHE2MA through model validation and simulation while demonstrating its ability to jointly reduce average service latency by 103.4% and energy consumption by 17.1% compared to a centralized RL solution.

As the demand for Network Function Virtualization accelerates, service providers are expected to advance the way they manage and orchestrate their network services to offer lower latency services to their future users. Modern services require complex data flows between Virtual Network Functions, placed in separate network domains, risking an increase in latency that compromises the offered latency constraints. This shift requires high levels of automation to deal with the scale and load of future networks. In this paper, we formulate the Service Function Chaining (SFC) placement problem and then we tackle it by introducing SCHEMA, a Distributed Reinforcement Learning (RL) algorithm that performs complex SFC orchestration for low latency services. We combine multiple RL agents with a Bidding Mechanism to enable scalability on multi-domain networks. Finally, we use a simulation model to evaluate SCHEMA, and we demonstrate its ability to obtain a 60.54% reduction of average service latency when compared to a centralised RL solution.

The standard for 5G communication exploits the concept of a network slice, defined as a virtualized subset of the physical resources of the 5G communication infrastructure. As a large number of network slices is deployed over a 5G network, it is necessary to determine the physical resource demand of each network slice, and how it varies over time. This serves to increase the resource efficiency of the infrastructure without degrading network slice performance. Traffic prediction is a common approach to determine this resource demand.State-of-the-art research has demonstrated the effectiveness of machine learning (ML) predictors for traffic prediction in 5G networks. In this context, however, the problem is not only the accuracy of the predictor, but also the usability of the predicted values to drive resource orchestration and scheduling mechanisms, used for resource utilization optimization while ensuring performance. In this paper, we introduce a new approach that consists on including problem domain knowledge relevant to 5G as regularization terms in the loss function used to train different state-of-the-art deep neural network (DNN) architectures for traffic prediction. Our formulation is agnostic to the technological domain, and it can obtain an improvement of up to 61,3% for traffic prediction at the base station level with respect to other widely used loss functions (MSE).

In cloud environments, the VMs within the computing nodes generate varying memory demand profiles. When memory utilization reaches its limits due to this, costly (virtual) disk accesses and/or VM migrations can occur. Since some nodes might have idle memory, some costly operations could be avoided by making the idle memory available to the nodes that need it. In view of this, new architectures have been introduced that provide hardware support for a shared global address space that, together with fast interconnects, can share resources across nodes. Thus, memory becomes a global resource. This paper presents a memory capacity aggregation mechanism for cloud environments called vMCA (Virtualized Memory Capacity Aggregation) based on Xen's Transcendent Memory (Tmem). vMCA distributes the system's total memory within a single node and globally across multiple nodes using a user-space process with high-level memory management policies. We evaluate vMCA using CloudSuite 3.0 on Linux and Xen. Our results demonstrate a peak running time improvement of 76.8% when aggregating memory, and of 37.5% when aggregating memory and implementing our policies.