Smart Connectivity is the next evolution in telecommunications, fully automated by AI. In this model, network services are designed and provisioned based on Service Level Agreements (SLAs). AI Native frameworks will control, manage, and optimize the AI/ML model pipelines that handle these service operations. This paper introduces a new intelligent orchestrator component, named NeuroPulse, that enables the integration of AI into the Virtual Network Functions (VNF) making them intelligent. This component monitors performance of deployed Virtual Network Intelligent Function (VNIFs), generates new configuration for VNIFs, and optimizes their performance. NeuroPulse is equipped with Few-Shot Profile Generator model that is capable of generating AI configurations aligned with the services’ SLA, and Performance Optimizer that attempts to suggest best previous model parameters for the generated configurations. The Few-Shot Profile Generator works in synergy with the Performance Estimator system. This estimator predicts the performance of generated profiles by analyzing historical data recorded by the monitoring system. The obtained results show that the NeuroPulse system can optimize the structure and the performance of the VNIFs to be aligned with the SLA terms agreed upon by the administrative domain and the service requester.

6G promises E2E cross domains continuous intelligence to optimize resource management and orchestration. However, state-of-the-art methods fall short in providing promised reliable and optimal resource management due to their inefficient proactive decision-making and planning capabilities. This paper proposes a novel Holistic Predictive Framework designed to enhance decision-making and achieve proactive multi-domain resource management. Our framework comprises of predictive, focus, and decision making elements, enabling exceptional proactive planning, and decisions-making based on a holistic vision of network's future. To select the best predictive and decisionmaking elements, various combinations of predictive Machine Learning (ML) and Reinforcement Learning (RL) algorithms were examined in our testbed. To demonstrate the superiority of our framework, we have conducted another test where our framework was compared with state-of-the-art solutions. The test results indicate that coupling the predictive element and attention-augmented decision making unit significantly improves the orchestrator's performance. Based on the result of both tests, our multi-domain orchestration solution, which exploits Latent ODE, outperforms all Cutting-Edge frameworks and is the best combination of the algorithms for our framework.

The rapid evolution towards 6G networks introduces new challenges in orchestrating services across distributed domains while ensuring sustainability goals, such as energy efficiency. Traditional scaling strategies focus on the number of network function instances without optimizing their placement based on energy consumption or resource usage. Addressing this gap, we propose a distributed and energy-efficient placement framework for scaled Network Function (NF) instances across multi-domain 6G infrastructures. Building upon a refined energy consumption model that accounts for computational and network-level power usage, we design a Graph Neural Network (GNN)-enhanced Deep Reinforcement Learning (DRL) agent to optimize placement decisions. The agent encodes the substrate topology and resource states to guide the selection of energy-efficient nodes during scaling operations. We implement and evaluate the framework in a realistic multi-domain scenario featuring fluctuating traffic patterns and heterogeneous node energy profiles. Results show that our approach reduces infrastructure energy consumption compared to a round-robin heuristic, while maintaining high placement success and efficient resource utilization. Among the DRL methods explored, Proximal Policy Optimization (PPO) achieved the best trade-off between placement stability, adaptability, and energy performance. These findings demonstrate the potential of GNN-based DRL agents for sustainable orchestration in future 6G networks.

The emergence of 6 G requires efficient management of heterogeneous networks and computational resources to achieve targeted end-to-end network automation, with slice orchestration as a key feature. Despite opportunities offered by the recent advances in network virtualization and distributed cloud infrastructures, these developments introduce complexity in the context of multi-domain networks. This paper presents a distributed horizontal scaling method that leverages deep reinforcement learning (DRL) to enhance network function orchestration (NFO) with intelligent scaling decisions, facilitating seamless cross-domain information exchange. Firstly, we develop a DRL agent designed to handle fluctuating traffic loads and generate scaling actions tailored for the considered network function (NF). The trained DRL agent is then integrated into a multi-domain message exchange scaling framework with traffic prediction capabilities. Moreover, a simulation testbed is developed to manipulate multi-domain topologies, customize network slices, and enable precise per-slice and per-domain scaling decisions. Our DRL-based solution outperforms the Horizontal Pod Autoscaling (HPA) heuristic used by Kubernetes, improving resource utilization and reducing data rate losses.

Network Data Analytics Function (NWDAF) is a 3GPP component that employs Machine Learning (ML) algorithms to analyze Network Functions (NFs) in the 5G Core. With the release of 3GPP Release 17, it is anticipated that the NWDAF could evolve to incorporate a distributed architecture, utilizing multiple edge-based components to facilitate parallel processing and thus reduce service latency. This paper presents a novel approach, Feedback-Driven Federated Learning (FDFL), for orchestrating distributed NWDAF architecture in the 5 G core. FDFL uses a feedback mechanism to guide ML models of client NWDAFs, ensuring faster convergence towardthe global function optima while preserving their data privacy. Additionally, this paper introduces Influence-based Weighted Federated Averaging to manage the impact of edge NWDAFs model parameters on global model aggregation. To enhance security, it is recommended to use Local Differential Privacy (LDP) for masking model parameter exchanges. Experimental results underscore the substantial superiority of our approach over other existing and state-of-the-art solutions.

The emergence of beyond 5G networks poses new challenges for network slicing orchestration. In particular, the distribution of orchestration mechanisms across multiple domains is a challenging problem, as independent entities must coordinate in a lightweight manner to maintain slice requirements. This paper presents a machine learning-driven horizontal scaling approach that uses inter-domain communication to achieve more accurate scaling decisions. We propose a collaborative orchestration approach based on the message-passing paradigm. Orchestration actions are performed using a Horizontal Pod Autoscaling (HPA) algorithm on each domain independently according to the required and allocated resources. We have developed a simulation testbed based on OMNeT++, allowing the creation and manipulation of multi-domain topologies and customized network slices. We show that exploiting inter-domain communication through message exchanges enables more accurate scheduling decisions than non-communicative orchestration, reducing data rate losses and enabling more efficient resource utilization.

The advanced 5G and 6G mobile network generations offer new capabilities that enable the creation of multiple virtual network instances with distinct and stringent requirements. However, the coexistence of multiple network functions on top of a shared substrate network poses a resource allocation challenge known as the Virtual Network Embedding (VNE) problem. In recent years, this NP-hard problem has received increasing attention in the literature due to the growing need to optimize resources at the edge of the network, where computational and storage capabilities are limited. In this demo paper, we propose a solution to this problem, utilizing the Algorithm Selection (AS) paradigm. This selects the most optimal Deep Reinforcement Learning (DRL) algorithm from a portfolio of agents, in an offline manner, based on past performance. To evaluate our solution, we developed a simulation platform using the OMNeT++ framework, with an orchestration module containerized using Docker. The proposed solution shows good performance and outperforms standalone algorithms.

Network slicing, also known as the virtual network embedding (VNE) problem, is an NP-hard optimization problem. Compared to traditional approaches, the methods relying on deep reinforcement learning yield better performance without exhibiting issues such as stacking at local minima and/or solutions’ space exploration limits. These algorithms present, however, different performances according to the employed approach, and the problem to be treated, resulting in robustness problems. To overcome these limits, we propose the adoption of the best algorithm, from a selection of learning strategies, in terms of reward and sample efficiency at each time step. The proposed strategy acts as a meta-algorithm that brings more robustness to the network by dynamically selecting the best solution for a specific scenario. Our solution proved its efficiency and managed to dynamically select the best algorithm in terms of the best acceptance ratio of the deployed services and outperform all the standalone algorithms.

Network slicing is one of the building blocks in Zero Touch Networks. It mainly consists in a dynamic deployment of services in a substrate network. However, the Virtual Network Embedding (VNE) algorithms used generally follow a static mechanism, which results in sub-optimal embedding strategies and less robust decisions. Some reinforcement learning algorithms have been conceived for a dynamic decision, while being time-costly. In this paper, we propose a combination of deep Q-Network and a Monte Carlo (MC) approach. The idea is to learn, using DQN, a distribution of the placement solution, on which a MC-based search technique is applied. This improves the solution space exploration, and achieves a faster convergence of the placement decision, and thus a safer learning. The obtained results show that DQN with only 8 MC iterations achieves up to 44% improvement compared with a baseline First-Fit strategy, and up to 15% compared to a MC strategy.

Network slicing remains one of the key technologies in 5G and beyond 5G networks (B5G). By leveraging SDN and NVF techniques, it enables the coexistence of several heterogeneous virtual networks (VNs) on top of the same physical infrastructure. Despite the advantages it brings to network operators, network slicing raises a major challenge: Resource allocation of VNs, also known as the virtual network embedding problem (VNEP). VNEP is known to be an NP-Hard problem. Several heuristics, meta-heuristics and Deep Reinforcement Learning (DRL) based solutions were proposed in the literature to solve it. Regarding the first two categories, they can provide a solution for large scale problems within a reasonable time, but the solution is usually suboptimal, which leads to an inefficient utilization of the resources and increases the cost of the allocation process. For DRL-based approaches and due to the exploration-exploitation dilemma, the solution can be infeasible. To overcome these issues, we combine, in this work, deep reinforcement learning and relational graph convolutional neural networks in order to automatically learn how to improve the quality of VNEP heuristics. Simulation results show the effectiveness of our approach. Starting with an initial solution given by the heuristics our approach can find an amelioration, with an improvement in the order of 35%.

Network Slicing (NS) is a key technology that enables network operators to accommodate different types of services with varying needs on a single physical infrastructure. Despite the advantages it brings, NS raises some technical challenges, mainly ensuring the Service Level Agreements (SLA) for each slice. Hence, monitoring the state of these slices will be a priority for ISPs. However, due to the high measurements overhead, it is generally forbidden to directly measure the performance of all of these slices. To overcome this limitation, network tomography is a promising solution, consisting of a set of methods of inferring unmeasured network metrics using end-to-end measurements between monitors. In this work, we focus on inferring the additive metrics of slices such as delays or logarithms of loss rates. We model the inference task as a regression problem that we solve using neural networks. In our approach, we train the model on an artificial dataset. This not only avoids the costly process of collecting a large set of labeled data but has also a nice covering property useful for the procedure's accuracy. Moreover, to handle a change on the topology or the slices we monitor, we propose a solution based on transfer learning in order to find a trade-off between the quality of the solution and the cost to get it. Simulation results with both, emulated and simulated traffic show the efficiency of our method compared to existing ones in terms of both accuracy and computation time.

Monitoring the status of network slices is a priority for network operators to ensure that SLAs are not violated. To overcome the limitations of direct slices’ monitoring, network tomography (NT) is seen as a promising solution. NT-based solutions require constraining monitoring traffic to follow specific paths, which we can achieve by using segment-based routing (SR). This allows deploying customized probing scheme, such as cycles’ probing. A major challenge with SR is, however, the limited length of the monitoring path. In this paper, we focus on the complexity of that task and propose MonGNN, a standalone solution based on Graph Neural Networks (GNNs) and genetic algorithms to find a trade-off between the quality of monitors’ placement and the cost to achieve it. Simulation results show the efficiency of our approach compared to existing methods.

The placement of Virtual Network Function - Forwarding Graphs (VNF-FGs) is one of the basic operations in the networks of the future. Being NP-hard, several heuristics and metaheuristics have been proposed. However, these approaches are inefficient due to the need to recalculate the solution at each service placement. In this paper, we adapt one of the most advanced approaches in Deep Reinforcement Learning (DRL), in order to improve exploration by generalizing the neural network calculating action values. We also propose an evolutionary algorithm to evolve these neural networks in order to discover better ones, which also avoids getting stuck in local minima. In order to avoid going through the almost innumerable number of infeasible solutions, we propose a heuristic, which combined with our DRL, makes it possible to guarantee the feasibility of the solutions and therefore to make the placement much more efficient. The simulation results we obtained confirm the quality of the solutions obtained as well as the superiority of the proposed solution over the existing one.

5G networks are expected to provide a variety of services over the same physical infrastructure equipped with a combination of radio and wired transport networks and leveraging network virtualization and Software Defined Networking (SDN). In particular, network slicing is envisioned as a promising solution to enable optimal support for heterogeneous services sharing the same infrastructure. To this end, we design and implement, in this paper, an SDN based architecture for end-to-end network slicing which proactively and dynamically adapts radio slices to the transport network slices. The developed architecture enables the creation, modification and continuity of radio and transport network slices while considering their resource and Quality-of-Service (QoS) requirements. It leverages Machine Learning for predicting radio slices capacities, improving network resource utilization, and predicting congestion within each network slice. We also formulate this network slicing problem as a Linear Program (LP) aiming to minimize the total network delay. Finally, we propose an efficient heuristic algorithm with low time complexity and high estimation accuracy to solve large problem instances. Experimental results using the OpenAirInterface (OAI) platform, FlexRAN, ONOS SDN Controllers and OpenvSwitch demonstrate the efficiency of our approach in terms of guaranteeing low latency and high network throughput.

Resource allocation of 5G network slices is one of the most important challenges for network operators. It can be formulated using the Virtual Network Embedding (VNE) problem, which was and remains an active field of studies, also known because of its NP-hardness. Owing to its complexity, several heuristics, meta-heuristics and Deep Learning-based solutions have been proposed. However, these solutions are inefficient either due to their slowness or to not taking into account the structure of data which results in an inefficient exploration of the solutions space. To overcome these issues, in this work we unveil the potential of Graph Convolutional Neural (GCN) networks and Deep Reinforcement Learning techniques in solving the VNE problem. The key point of our approach is modeling of the VNE problem as an episodic Markov Decision Process which is solved in a Reinforcement Learning fashion using a GCN-based neural architecture. The simulation results highlight the efficiency of our approach through an increased performance over time, while outperforming state-of-art solutions in terms of the services' acceptance ratio.

The centralization of network intelligence enabled by Software Defined Networking (SDN), and the recent breakthroughs of Machine Learning (ML), paved the way to address a variety of network challenges. Quality-of-Service (QoS)-aware routing is one of the important challenges in SDN-based networks, especially when multiple flows coexist in the same network. Optimizing network performances (i.e., end-to-end delay, throughput) must be achieved in order to enhance the performance of QoS-aware routing. Neural Networks and Reinforcement learning are ML breakthroughs that can tackle this important challenge. To this end, we propose, in this paper, an efficient rules placement algorithm based on Deep Reinforcement Learning (DRL) and traffic prediction. Our proposal aims to dynamically collect the optimal path from the DRL agent and predict future traffic demands using the well-known prediction method LSTM. To do so, we first formulate the flow rules placement as an Integer Linear Program (ILP) that aims to minimize the total network delay. Then, we propose a simple yet efficient heuristic algorithm to solve the formulated ILP problem with low time complexity and high estimation accuracy. The proposed algorithm interacts with the DRL agent to get the optimal path and the traffic prediction module in order to avoid congestion. The obtained results using ONOS controller and OpenvSwitch revealed the efficiency of the proposed approach in decreasing both network latency and packet loss, and rising the network throughput.

Network Function Virtualization (NFV) and service orchestration simplify the deployment and management of network and telecommunication services. The deployment of these services requires, typically, the allocation of Virtual Network Function - Forwarding Graph (VNF-FG), which implies not only the fulfillment of the service's requirements in terms of Quality of Service (QoS), but also considering the constraints of the underlying infrastructure. This topic has been well-studied in existing literature, however, its complexity and uncertainty of available information unveil challenges for researchers and engineers. In this paper, we explore the potential of reinforcement learning techniques for the placement of VNF-FGs. However, it turns out that even the most well-known learning technique is ineffective in the context of a large-scale action space. In this respect, we propose approaches to find out feasible solutions while improving significantly the exploration of the action space. The simulation results clearly show the effectiveness of the proposed learning approach for this category of problems. Moreover, thanks to the deep learning process, the performance of the proposed approach is improved over time.

In Software Defined Networking, due to the significant bandwidth and latency requirements, predicting available resources (i.e., latency, bandwidth) is crucial for enhancing performances, resource utilization and power consumption of the data plane. In this paper, we propose an efficient rules placement algorithm based on predictive network latency using online learning. Our proposal aims to dynamically predict the latency for updating the flow rules in network devices. To do so, we first formulate the flow rules placement as an Integer Linear Program (ILP) that aims to minimize the total network delay. Then, we propose a simple yet efficient heuristic algorithm to solve the formulated ILP problem with low time complexity. Experimental results using ONOS controller and Mininet show the efficiency of our proposal in decreasing network latency, packet loss and rising the network throughput.

Future mobile networks (5G) would be much more flexible, dynamic and faster adaptable to match the evolved demand. Network Function Virtualization (NFV) is investigated to take the advantage of information technology (IT) virtualization on telecom networks by separating network functions to be virtualized from underlying dedicated hardwares. Container-based micro-service is currently discussed as being lightweight virtualization approach enabling flexibility and scalability of future mobile networks. Virtualizing mobile network functions, the Serving Gateway (SGW) is analyzed as being the most sensible component, a bottleneck. The containerization seems to be the adequate approach to overcome this issue as it could enable rapid deployment by scaling SGW instances based on workload. In this paper, we discuss the cloudification of mobile network functions using containerization technology and 12 factors principles for enabling such cloudification. An implementation is made using Docker containerization while we employ Docker Swarm for cluster management and deployment. We also build an orchestration testbed for testing the scalability of SGW. This is realized by comparing the performances of two open-source orchestrators: Kubernetes and Mesos- Marathon. The proof of concepts shows clearly that a container-based approach is a viable option for achieving elasticity of future mobile networks.

The introduction of micro-services in cloud infrastructure provides modularity, flexibility and distributed software components. In telecom, the Network Function Virtualization enables running an application as a collection of smaller software components and micro-services that share an operating system (OS) or distributed across many computing servers in the cloud. In this paper, we present a micro-service platform to target the flexibility of telecom networks and the automation of its deployment. Using Docker orchestration, this demo paper shows the flexibility and the rapid deployment of a wireless network infrastructure.

Automatic video editing is a hot topic due to the rapid growth of video usages. In this paper, we present a cloud-based tool and an approach of automatic video editing based on keywords extracted from the audio transcription. Using texts transcript of the audio, video sequences are selected and chained to create automatically a new video with a time duration fixed by the user. A cloud based video editor allows users to collaboratively edit the video.