To address the challenges posed by the deployment of microservices of future end-user applications in the cloud continuum, a performance prediction model working together with a network elasticity controller will be needed. With that aim, this work introduces Layered Queuing-Graph Neural Networks (LQ-GNN), a novel Machine Learning (ML) approach to develop a generalized performance prediction model for microservicebased applications. Unlike previous works focused on individual applications, our proposal aims for a versatile model applicable to any microservice-based application, integrating the Layered Queueing Network (LQN) modeling with Graph Neural Networks (GNN). LQ-GNN allows to efficiently estimate the response time of applications under different resource allocations and placements on the computing continuum. The obtained evaluation results indicate that the proposed model achieves a prediction error below 10% when considering different evaluation scenarios. Compared to existing methodologies, our approach balances prediction accuracy and computational efficiency, making it viable for real-time deployments. Consequently, ML-based performance prediction can significantly enhance the resource management and elasticity control of microservice-based architectures, leading to more resilient and efficient systems.

Future applications to be supported by 6G networks are envisaged to be realized by loosely-coupled and independent microservices. In order to achieve an optimal deployment of applications, smart resource management strategies will be required, working in a cost-effective and resource-efficient manner. Current cloud computing services are challenged to meet the explosive growth and demand of future use cases such as virtual/augmented/mixed reality (VR/AR/MR). The purpose of edge computing (EC) is to better address latency and transmission requirements of those future stringent applications. However, a high flexibility and a rapid decision-making will be required since EC suffers from limited resources availability. For this reason, this work proposes an artificial intelligence (AI) technique, based on reinforcement learning (RL), to make intelligent decisions on the optimal tier and edge-site selection to serve any request according to the application’s category, constraints, and conflicting costs. In addition, when deployed at the edge-network, a heuristic has been proposed for the mapping of microservices within the selected edge-site. That heuristic will exploit a ranking methodology based on the network topology and available network and compute resources while preserving the revenue of the mobile network operator (MNO). Simulation results show that the performance of the proposed RL approach is close to the optimal solution by reaching the cost minimization objective within a 8.3% margin; moreover, RL outperforms considered benchmark algorithms in most of the conducted experiments.

Cognizant of the ease with which softwarized functions can be dynamically scaled according to real time resource requirements, and the fact that multiple services can have common VNFs in their chaining, this paper tackles the problem of cost effective deployment of online services from the perspective of sharing their VNF instances. First, we formally formulate the deployment problem under VNFs sharing. Secondly, given the NP-hard nature of the above problem, we propose a reinforcement learning (RL) algorithm capable of making intelligent placement decisions while considering multiple conflicting costs. Costs of transmission, VNF instantiation or energy consumption, among others. Thanks to the intelligence of the RL algorithm, simulation results show that the performance of the proposed algorithm is within a 14% margin and similar to an optimal solution in terms of request provisioning cost and acceptance ratio, respectively. Moreover, the algorithm results in more than a 20% and a 70% improvement in terms of request deployment cost and time compared to a state-of-the-art algorithm, and up to more than a 40% improvement in terms of cost compared to an algorithm that greedily minimizes the transmission or VNF activation costs.

To support future 5G computing and communication scenarios, cloud-network management tools should deploy cloud-network services adopting uncomplicated ways, reducing not only the time to market but also broadening the community capable of deploying new services. In this paper, we present the support of NECOS Platform, an EU-Brazil jointly funded project, towards slice-as-a-service creation from non-structured service description. We describe how NECOS architecture allows such functionality during the slice creation loop, and we present the initial efforts we took for structuring such a mechanism.

Network slicing has recently been proposed as one of the main enablers for 5G networks; it is bound to cope with the increasing and heterogeneous performance requirements of these systems. To "slice" a network is to partition a shared physical network into several self-contained logical pieces (slices) that can be tailored to offer different functional or performance requirements. Moreover, a defining characteristic of the slicing paradigm is to provide resource isolation as well as efficient use of resources. In this context, the thesis described in this paper contributes to the problem of slicing WiFi networks by proposing a solution to the problem of enforcing and controlling slices in WiFi Access Points. The focus of the research is on a variant of network slicing called QoS Slicing, in which slices have specific performance requirements. In this document, we describe the two main contributions of our research, a resource allocation mechanism to assign resources to slices, and a solution to enforce and control slices with performance requirements in WiFi Access Points.

Network slicing has recently been proposed as one of the main enablers for 5G networks. The slicing concept consists of the partition of a physical network into several self-contained logical networks (slices) that can be tailored to offer different functional or performance requirements. In the context of 5G networks, we argue that existing ubiquitous WiFi technology can be exploited to cope with new requirements. Therefore, in this paper, we propose a novel mechanism to implement network slicing in WiFi Access Points. We formulate the resource allocation problem to the different slices as a stochastic optimization problem, where each slice can have bit rate, delay, and capacity requirements. We devise a solution to the problem above using the Lyapunov drift optimization theory, and we develop a novel queuing and scheduling algorithm. We have used MATLAB and Simulink to build a prototype of the proposed solution, whose performance has been evaluated in a typical slicing scenario.

Network Function Virtualization (NFV) and Machine Learning (ML) are envisioned as possible techniques for the realization of a flexible and adaptive 5G network. ML will provide the network with experiential intelligence to forecast, adapt and recover from temporal network fluctuations. On the other hand, NFV will enable the deployment of slice instances meeting specific service requirements. Moreover, a single slice instance may require to be deployed across multiple substrate networks; however, existing works on multi-substrate Virtual Network Embedding fall short on addressing the realistic slice constraints such as delay, location, etc., hence they are not suited for applications transcending multiple domains. In this paper, we address the multi-substrate slicing problem in a coordinated manner, and we propose a Reinforcement Learning (RL) algorithm for partitioning the slice request to the different candidate substrate networks. Moreover, we consider realistic slice constraints such as delay, location, etc. Simulation results show that the RL approach results into a performance comparable to the combinatorial solution, with more than 99% of time saving for the processing of each request.

In forthcoming 5G networks, slicing has been proposed as a means to partition a shared physical network infrastructure into different self-contained logical parts (slices), which are set up to satisfy certain requirements. Although the topic has been thoroughly investigated by the scientific community and the industry, there are not many works addressing the challenges that appear when trying to exploit slicing techniques over WiFi networks. In this paper, we propose a novel method of allocating resources for WiFi networks to satisfy minimum bit rate requirements. We formulate an optimization problem, and we propose a solution based on the theory of Lyapunov drift optimization. The validity of the proposed solution is assessed by means of a simulation-based evaluation in Matlab.

Network slicing has recently appeared as one of the most important features that will be provided by 5G networks and is attracting considerable interest from industry and academia. At the wireless edge of these networks, most of the contributions in this area are related to cellular technologies leaving behind WiFi networks. In this work, we present a resource allocation mechanism based on airtime assignment to achieve infrastructure sharing and slicing in WiFi Access Points. The approach is simple and has the potential to be straightforwardly used within scenarios of wireless access infrastructure sharing.

New architectural and design approaches for radio access networks have appeared with the introduction of network virtualization in the wireless domain. One of these approaches splits the wireless network infrastructure into isolated virtual slices under their own management, requirements, and characteristics. Despite the advances in wireless virtualization, there are still many open issues regarding the resource allocation and isolation of wireless slices. Because of the dynamics and shared nature of the wireless medium, guaranteeing that the traffic on one slice will not affect the traffic on the others has proven to be difficult. In this paper, we focus on the detailed definition of the problem, discussing its challenges. We also provide a review of existing works that deal with the problem, analyzing how new trends such as software defined networking and network function virtualization can assist in the slicing. We will finally describe some research challenges on this topic.

The virtualization of Radio Access Networks (RANs) has been proposed as one of the important use cases of Network Function Virtualization (NFV). In Virtualized Radio Access Networks (VRANs), some functions from a Base Station (BS), such as those which make up the Base Band Unit (BBU), may be implemented in a shared infrastructure located at either a data center or distributed in network nodes. For the latter option, one challenge is in deciding which subset of the available network nodes can be used to host the physical BBU servers (the placement problem), and then to which of the available physical BBUs each Remote Radio Head (RRH) should be assigned (the assignment problem). These two problems constitute what we refer to as the VRAN Placement and Assignment Problem (VRAN-PAP). In this paper, we start by formally defining the VRAN-PAP before formulating it as a Binary Integer Linear Program (BILP) whose objective is to minimize the server and front haul link setup costs as well as the latency (or distance) between each RRH and its assigned BBU. Since the BILP could become computationally intractable, we also propose a greedy approximation for larger instances of the VRAN-PAP.

In this paper we propose the use of a two-agent learning scheme for the management of radio resources on cellular access networks. The management is materialized by the implementation of a self-configuration system governing the setup of several parameters on each base station. The two agents have independent goals; one is trying to maximize the quality of service and the other the economic benefit. Thanks to the combined use of the fuzzy logic technique and reinforcement learning, both agents will work in a complementary mode, achieving both goals simultaneously.

Network virtualisation is a promising technique for dealing with the resistance of the Internet to architectural changes. This is achieved by enabling a novel business model in which infrastructure management is decoupled from service provision. One of the main challenges in network virtualisation is efficient sharing of physical network resources by the different virtual networks. This work contributes to efficient resource sharing in network virtualisation by dividing the resource management problem into three sub-problems: virtual network embedding (VNE), dynamic resource allocation (DRA), and virtual network survivability (VNS); and then proposing a solution for each one of them. Specifically, we propose a path generation-based approach for VNE, machine learning-based self-management approaches for DRA, and a multi-entity negotiation algorithm for VNS. Through simulations, all our proposals are compared with related approaches, showing improvements in resource utilisation efficiency, which would directly result into better profitability for physical resource owners.

Network function virtualization has received attention from both academia and industry as an important shift in the deployment of telecommunication networks and services. It is being proposed as a path towards cost efficiency, reduced time-to-markets, and enhanced innovativeness in telecommunication service provisioning. However, efficiently running virtualized services is not trivial as, among other initialization steps, it requires first mapping virtual networks onto physical networks, and thereafter mapping and scheduling virtual functions onto the virtual networks. This paper formulates the online virtual function mapping and scheduling problem and proposes a set of algorithms for solving it. Our main objective is to propose simple algorithms that may be used as a basis for future work in this area. To this end, we propose three greedy algorithms and a tabu search-based heuristic. We carry out evaluations of these algorithms considering parameters such as successful service mappings, total service processing times, revenue, cost etc, under varying network conditions. Simulations show that the tabu search-based algorithm performs only slightly better than the best greedy algorithm.

As the virtualization of networks continues to attract attention from both industry and academia, the virtual network embedding (VNE) problem remains a focus of researchers. This paper proposes a one-shot, unsplittable flow VNE solution based on column generation. We start by formulating the problem as a path-based mathematical program called the primal, for which we derive the corresponding dual problem. We then propose an initial solution which is used, first, by the dual problem and then by the primal problem to obtain a final solution. Unlike most approaches, our focus is not only on embedding accuracy but also on the scalability of the solution. In particular, the one-shot nature of our formulation ensures embedding accuracy, while the use of column generation is aimed at enhancing the computation time to make the approach more scalable. In order to assess the performance of the proposed solution, we compare it against four state of the art approaches as well as the optimal link-based formulation of the one-shot embedding problem. Experiments on a large mix of virtual network (VN) requests show that our solution is near optimal (achieving about 95% of the acceptance ratio of the optimal solution), with a clear improvement over existing approaches in terms of VN acceptance ratio and average substrate network (SN) resource utilization, and a considerable improvement (92% for a SN of 50 nodes) in time complexity compared to the optimal solution.

Network virtualisation is considerably gaining attention as a solution to ossification of the Internet. However, the success of network virtualisation will depend in part on how efficiently the virtual networks utilise substrate network resources. In this paper, we propose a machine learning-based approach to virtual network resource management. We propose to model the substrate network as a decentralised system and introduce a learning algorithm in each substrate node and substrate link, providing self-organization capabilities. We propose a multiagent learning algorithm that carries out the substrate network resource management in a coordinated and decentralised way. The task of these agents is to use evaluative feedback to learn an optimal policy so as to dynamically allocate network resources to virtual nodes and links. The agents ensure that while the virtual networks have the resources they need at any given time, only the required resources are reserved for this purpose. Simulations show that our dynamic approach significantly improves the virtual network acceptance ratio and the maximum number of accepted virtual network requests at any time while ensuring that virtual network quality of service requirements such as packet drop rate and virtual link delay are not affected.

Approaches based on signal-strength measurements are at present the most popular on indoor location due to their reasonable accuracy and cost effective deployment. In this paper, we present an outlook of our research on different alternatives to implement an indoor location approach based on signal-strength measurements for an uncoordinated environment, an environment where we do not have any control on the number of access points, their location, availability or transmitted power.

The so-called second generation P2P file-sharing applications have with no doubt a better performance than the first implementations. The most remarkable difference is due to the file division into smaller pieces, where a receiving peer of any piece automatically becomes a new source to other peers. But a new question arises on how we distribute all the pieces provided by a seed peer to minimize the global and presumably individual download times. In this paper we summarize part of the work we have developed up until now to answer this general question, in particular, we will analyze how close the present second generation P2P file-sharing applications remain from an ideal solution with the theoretical best performance, that is, where all peers are interconnected with each other and all peers have an altruistic behavior always uploading its contents at any chance. Successive modifications of the ideal solution will lead us to more realistic scenarios. We will estimate the performance on each case and finally present the current studies we are carrying out to improve the overall capacity.