Streaming 360° video is challenging, as its bandwidth exceeds that of most users. A popular solution to this problem is tiled-based streaming. Live streaming of tiled 360° video requires encoding tiles under strict time constraints. Hardware encoders, such as NVIDIA's NVENC, are used to meet this requirement. However, selecting appropriate encoding parameters for NVENC is difficult because existing benchmarks do not consider tiled-based streaming and evaluate only a subset of the available encoding parameters. In this paper, we benchmark the rate-distortion performance and the encoding speed of NVENC and recommend encoding parameters for live 360° video tiled streaming. Additionally, since we deem unfit existing 360° videos for streaming use cases, we first craft a dataset of twenty long-duration 8K/30 fps stereoscopic 360° videos, provided in both stitched and unstitched formats. Our code and data is available publicly at https://github.com/Tiled360Benchmark/360DatasetAndBenchmark.

• Developed Monitoring tool collectors for fine-grained edge workload monitoring • Lightweight pipeline for intent-based assurance on resource-constrained devices • Real-time detection of application transitions using an autoencoder model • Fast and accurate Application Type Identification via Random Forest classifier • Public AIMED-2025 dataset with 9 workloads on RPi to support the research community Intent-Based Networking (IBN) enables agile and policy-driven network management by translating high-level intents into concrete configurations and continuously validating their compliance. A critical limitation in current Intent-Based Network Assurance (IBNA) systems is the lack of real-time application-level awareness, particularly in dynamic edge environments where AI workloads frequently change. In this work, we address this limitation by introducing a lightweight, monitoring-driven pipeline that enables the detection of application transitions and identification of newly active application types on edge devices. In collaboration with Netdata engineers, we develop multimetric data collectors using Netdata, an open-source platform for real-time system and application monitoring. These collectors capture application-agnostic system metrics with minimal overhead, forming the foundation for real-time alerting and dynamic network adaptation. Our proposed pipeline transforms raw monitoring data into fixed-length vectorized multivariate time series. An undercomplete autoencoder is then used to detect changes in system behavior indicative of application transitions, followed by a Random Forest classifier that labels the newly active application based on its resource usage profile. To support reproducibility, we construct and publicly release the AIMED-2025 dataset, which includes monitoring data from seven MediaPipe-based edge AI applications and two idle states, all executed on a Raspberry Pi. Experimental evaluation demonstrates that our method achieves 100% accuracy in both Application Transition Detection and Application Type Identification using only a three-second observation window. Furthermore, the system exhibits sub-second training times and millisecond-scale inference latency, making it suitable for real-time deployment on resource-constrained edge devices. Once an application change is detected and identified, the IBNA system can automatically alert network administrators and trigger dynamic reconfiguration of network resources to meet the specific performance, security, and connectivity requirements of the active application. By integrating application-level awareness into IBNA, this work advances the state of the art in intent-driven network management and enables more adaptive, efficient, and reliable operation of edge AI systems.

Radio Link Failure (RLF) detection is essential for maintaining reliable connectivity in 5G networks. However, traditional centralized detection mechanisms often encounter scalability and latency constraints when managing large-scale, geographically distributed infrastructures. To address this challenge, we introduce a Lyapunov-driven federated learning framework that adaptively selects gNodeBs based on both data utility and historical participation. This approach leverages an LSTM-based local model to capture temporal patterns in link performance, thereby enhancing RLF detection. Extensive evaluations on a real-world 5G dataset demonstrate that the proposed method achieves superior performance compared to baseline approaches when detecting rare failure events. By simultaneously prioritizing performance and fairness, this framework offers a scalable solution suited to diverse and dynamic 5G environments.

Unmanned Aerial Vehicles (UAVs) are expanding beyond military use into sectors such as logistics, communication, and transportation. However, their dependence on high-power batteries limits their range, and while fuel cells provide longer flight times, they can reduce speed and acceleration due to safety concerns. Consequently, managing UAV power consumption has become a critical challenge, directly affecting flight duration and operational performance. Therefore, accurately predicting power consumption is important for enabling efficient UAV mission planning. Thus, in this paper, we propose a fine-tuning strategy called Large Language Model for Power Forecasting (LLM4PF) that employs a Generative Pretrained Transformer (GPT-2) model to reduce computational costs without sacrificing accuracy. LLM4PF predicts power consumption based on various UAV operational data including speed and altitude among others. Furthermore, we evaluate its performance in low-data scenarios through few-shot learning with 5% and 10% data subsets. Additionally, we compare LLM4PF to transformer-based models using a public dataset, demonstrating its effectiveness and efficiency.

The rapid expansion of Artificial Intelligence (AI) applications at the Edge has created an increasing demand for energy-efficient deployment. Furthermore, Edge devices are inherently constrained in computation, bandwidth, and storage, making the exploration of compressed AI models and data a worthwhile approach to enhancing efficiency. While compression improves resource and energy efficiency, it often leads to significant performance degradation, especially with complex data. To address this challenge, we propose leveraging transformations on compressed data to enhance the effectiveness of compressed AI models. We evaluate 13 different transformations across three benchmark datasets (MNIST, FashionMNIST, CIFAR-10) and find that applying shadow transformations to lossy compressed images significantly mitigates performance loss. Based on experiments in a real edge device our approach achieves a 10.7% reduction in energy consumption, 99.48% compression in AI model architecture, and up to 72% data size reduction, with performance degradation ranging only from 0.14% to 0.89%. These results show that data transformations can enable efficient Edge AI inference with minimal performance loss, reducing energy, bandwidth, and computation.

Modern networks are dynamic and heterogeneous, leading to significant challenges for Intrusion Detection Systems (IDS) due to data drift and concept drift phenomenons. Traditional Machine Learning (ML) models trained on a source domain often suffer performance degradation when deployed in a target domain, while online learning models require substantial data to adapt effectively. To address these limitations, we propose a hybrid methodology that integrates online learning with domain adaptation for intrusion detection (OLIDA-IDS). OLIDA-IDS begins with a static random forest model trained on the source domain, employs a Stacked Marginalized Denoising Autoencoder (sMDA) for unsupervised domain adaptation to align feature distributions, and transitions to an Adaptive Random Forest (ARF) for online learning as target data becomes available incrementally. Experimental results demonstrate that OLIDA-IDS achieves 99.59% accuracy in the target network environment, outperforming static, semi-supervised, and incremental learning approaches. Key contributions include the novel integration of sMDA for domain adaptation, a weighted-accuracy technique for seamless transition between static and online models, and the use of ARF for superior resilience in evolving network threats. This work bridges the gap between domain adaptation and online learning, offering a feasible solution for real-world IDS deployment in dynamic environments.

Radio Link Failure (RLF) detection in Radio Access Networks (RANs) is crucial for ensuring seamless communication in 5G networks. Nonetheless, current approaches based on traditional Machine Learning (ML) algorithms fail to find a trade-off between accuracy and computational complexity. Thus, in this paper, we explore advanced and computationally efficient types of transformers, such as Linformer and Performer. These models significantly reduce computational complexity while maintaining strong performance by leveraging different attention mechanisms. Extensive evaluations using a realistic dataset under various percentages of link failures show that Linformer provides a favourable balance between accuracy and training time.

The increasing demand for compute intensive Internet of Thing (IoT) applications has accelerated the adoption of multi-access edge Computing (MEC) to offload tasks from resource constrained devices to edge servers. However, making optimal offloading decisions in multi-user MEC environments is challenging due to the dependencies between tasks, resource constraints, and the need to preserve user privacy. In this work, we propose FEDORA, a federated ensemble reinforcement learning framework for directed acyclic graph (DAG)-based task Offloading and resource allocation in MEC environments, that integrates twin delayed deep deterministic policy gradient (TD3) for continuous resource allocation and multi-head deep Q-networks (DQN) for discrete offloading decisions. To handle task dependencies, we model applications as DAGs and generate feature embeddings for offloading decisions. Our federated learning (FL) approach uses local training at MEC level and periodic model aggregation at a global server to preserve data privacy. Finally, extensive simulations across different DAG topologies demonstrate that FEDORA reduces system costs and improves task completion rates compared to state-of-the-art baselines including FL-DQN, FL-DDPG, FedAvg, FedNova, and SCAFFOLD, highlighting its scalability and robustness in large scale MEC deployments.

Intrusion Detection systems (IDS) are alerting cybersecurity tools that analyze network traffic in order to identify suspicious activity and known threats. State of the art IDS rely on supervised machine learning models which are trained to categorize the network flow with a historical labeled dataset. Nonetheless, next-generation networks are characterized as heterogeneous and dynamic. The heterogeneity can make every network environment to be significantly different and the dynamicity means that new threats are constantly emerging. These two factors raise the research question if a supervised machine learning based IDS can work efficiently in a network environment different from the one that generated its labeled training data. In this paper, we first give an answer to this research question and next try to propose a semi-supervised learning approach that can be generalized sufficiently in a different network environment using unlabeled data, taking into consideration that unlabeled data are much easier and cheap to be collected compared to labeled ones. In order to have a proof of concept we made experiments with two labeled datasets CIC-IDS2017, CIC-IDS2018 which are publicly available and one unlabeled dataset PS-Azure2023 which we constructed for this work and make it also publicly available. The results confirm our assumption and the applicability of the semi-supervised learning paradigm for the design of IDS.

Internet of Things (IoT) applications generate tremendous amounts of data streams which are characterized by varying Quality of Service (QoS) indicators. These indicators need to be accurately estimated in order to appropriately schedule the computational and communication resources of the access and Edge networks. Nonetheless, such types of IoT data may be produced at irregular time instances, while suffering from varying network conditions and from the mobility patterns of the edge devices. At the same time, the multipurpose nature of IoT networks may facilitate the co-existence of diverse applications, which however may need to be analyzed separately for confidentiality reasons. Hence, in this paper, we aim to forecast time series data of key QoS metrics, such as throughput, delay, packet delivery and loss ratio, under different network configuration settings. Additionally, to secure data ownership while performing the QoS forecasting, we propose the FeDerated Temporal Sparse Transformer (FeD-TST) framework, which allows local clients to train their local models with their own QoS dataset for each network configuration; subsequently, an associated global model can be updated through the aggregation of the local models. In particular, three IoT applications are deployed in a real testbed under eight different network configurations with varying parameters including the mobility of the gateways, the transmission power and the channel frequency. The results obtained indicate that our proposed approach is more accurate than the identified state-of-the-art solutions.

In this paper we address the challenge of efficiently deploying Virtual Network Functions (VNFs) in network infrastructures. This is particularly crucial when facing resource fragmentation, where available resources are not fully utilized due to the fluctuating allocation and deallocation of virtual network requests. Traditional optimization techniques often fall short in managing the dynamic complexities of VNF placement. To overcome this, we introduce a novel online VNF placement strategy using Deep Reinforcement Learning (DRL) combined with a Reward Constrained Policy Optimization (RCPO). This method leverages the flexibility of DRL and the constraint integration capacity of RCPO, ensuring compliance with performance and resource limitations while minimizing resource fragmentation. The results demonstrate that our DRL-based method surpasses existing methods, resulting in more effective resource management and less resource fragmentation.

Next generation networks will be largely based on monitoring and telemetry tools that are essential for maintaining optimal performance, ensuring security, managing costs, and performing fault detection and resolution. An integral part of the overall monitoring strategy is alerting, which provides administrators with the necessary information to proactively or reactively manage and optimize network services. However, when monitoring systems generate an excessive number of alerts, many of which may not be actionable or may not represent critical issues, the phenomenon of alert fatigue occurs. Alert fatigue refers to a situation where the volume and the speed of the continuous influx of alerts becomes so overwhelming that the network administrators become desensitized and do not respond to them. To this end, and inspired by recent trends in network automation, where human intervention tends to be minimized, we introduce an alert fatigue mitigation mechanism in monitoring focusing on cloud computing infrastructures. In particular, a composite machine learning methodology is proposed in order to select which alerts will be hidden and which ones will be presented to the administrators. Additionally, to personalize the results, the proposed approach considers the level of users’ experience along with the alert features to further optimize the accuracy of the alert filtering mechanism. The research has been conducted in a realistic environment of a leading monitoring enterprise, Netdata, which provided two datasets for testing our approach. Furthermore, the attained results of the filtering mechanism were evaluated by expert engineers of the company that verified the output of the proposed framework. Specifically, the outcomes confirm that our proposed methodology mitigates the alert fatigue problem with an accuracy that surpass 90% in most cases.

In this paper we examine different approaches for the prediction of Service Level Agreements (SLAs) violations that occur during the service provisioning between cloud customers and providers. Despite the fact that there are many network metrics that involve the server - client interaction, it is an open research question how these available metrics can be used by a SLA prediction mechanism. We study three different data representation models for the network characteristics, a time series, a content and a context representation. We see that a context approach using graph representations captures efficiently the associativity of clients and improves the performance of traditional SLA violation prediction models when it is combined with them. The prediction of the SLA violations takes place using neural networks, making us propose a composite SLA prediction model that leverages Graph Neural Networks (GNNs). In our research, we put special emphasis and try different variations on how we construct the graphs. We perform an extensive performance evaluation of 23 different SLA prediction models that can be grouped into the three representations categories, namely the vector models that are based on network features, sequential models that leverage the temporal evolution of QoS metrics and Graph models that take into consideration the associativity of the clients. The experimental results show that our proposed GNN-based model can significantly improve the accuracy of SLA violation prediction, making it a useful tool for Cloud and Service providers.

The recent networking trends driven primarily by the different virtualization technologies, such as Network Function Virtualization (NFV) and Service Function Chaining (SFC) pave the way for next-generation network services. In the 5G and beyond era, such services usually have strict delay requirements and the wider adoption of the distribution of their computational needs across the Edge-to-Cloud continuum is certainly a step in the right direction. However, the majority of the optimization solutions for placing the virtualized services so far focus on server selection, leaving other areas such as the impact of Non-Uniform Memory Access (NUMA) and CPU core selection underexplored. In this work, we herein formulate the problem of placing services as SFCs on an Edge/Cloud infrastructure, as a Mixed Integer Programming (MIP) problem. Then, we propose a heuristic algorithm called “Dynamic numa node Selection through Cores consolidation – DySCo" to solve it, which optimizes the placement in terms of server, NUMA and core selection. To the best of our knowledge, this is the first attempt to optimize network service placement in an Edge-Cloud interplay. Extensive simulation evaluation shows that DySCo is able to perform close to optimal while finding a solution in a real time fashion. Compared to a mix of baselines and modified solutions from the literature to treat this new problem, DySCo reduces on average the deployment cost by 17.53% and the delay by 28.88% for a given SFC.

Low-earth-orbit satellite networks (LSNs) offer an enhanced global connectivity and a wide range of applications such as disaster response and military operations, among others. Each specific application can be represented by a service function chain (SFC) in which each function is considered as a task in the application. Our objective is to optimize the long-term system performance by minimizing the average end-to-end delay of SFC deployments in LSNs. To achieve this, we formulate a dynamic programming (DP) problem to derive an optimal placement policy. To overcome the computational intractability, the need for statistical knowledge of SFC requests, and centralized decision-making challenges, we present a multi-agent Q-learning approach where satellites act as independent agents. To facilitate performance convergence in non-stationary agents' environments, we let agents to collaborate by sharing designated learning parameters. In addition, agents update their Q-tables via two distinct rules depending on selected actions. Extensive experimentation shows that our approach achieves convergence and performance relatively close to the optimum obtained by solving the formulated DP equation.

In the 5G/6G era of networking, computational offloading, i.e., the act of transferring resource-intensive computational tasks to separate external devices in the network proximity, constitutes a paradigm shift for mobile task execution on Edge Computing infrastructures. However, in order to provide firm Quality of Service (QoS) assurances for all the involved users, meticulous planning of the offloading decisions should be made, which potentially involves inter-site task transferring. In this paper, we consider a multi-user, multi-site Multi-Access Edge Computing (MEC) infrastructure, where mobile devices (MDs) can offload their tasks to the available edge sites (ESs). Our goal is to minimize end-to-end delay and energy consumption, which constitute the sum cost of the considered system, and comply with the MDs’ application requirements. To this end, we introduce a two-stage Reinforcement Learning (RL)-based mechanism, where the MDs-to-ES task offloading and the ES-to-ES task transferring decisions are iteratively optimized. The proper operation, effectiveness and efficiency of our proposed offloading mechanism is assessed under various evaluation scenarios.

The unprecedented expansion of network communications and the addition of new applications and services have made the network configuration a riddle for network engineers. At the same time, the daily interaction of the users with various network applications and the heterogeneous requirements they impose, necessitates a more direct communication between these two entities. Intent Based Networking (IBN) is a new technology that enables this by automating the network configuration, while allowing the users to directly interact with the network. The latter can be established through a natural language form, letting the users to freely express their high-level network requirements via a normal conversational approach. To this end, in this paper, we introduce ETS-Chatbot by extending the functionalities of a chatbot in order to be used in an IBN context. We position our work in an enterprise environment and we try to capture the high-level intent and map it into a model that can be understood by the network. The obtained results reveal that our framework can accurately classify the intent of a user with high precision, giving a strategic advantage to the enterprise.

Cloud gaming offers excellent potential; however, it presents significant challenges for 5G networks because of its strict requirements for high reliability and low latency. Cloud gaming service deployment can be modeled as a Virtual Network Functions Chain Placement Problem (VNF-CPP), where a service instance is represented by a chain of interconnected Virtual Network Functions (VNFs). Solving the VNF-CPP may result in infeasible solutions when the underlying network infrastructure can not meet the service requirements. In this paper, we propose an automatic feasibility restoration technique and explain how network operators can use it to meet cloud gaming's demands. Our proposed algorithms reveal the origins of the infeasibilities so that network operators can understand the reasons behind them. Moreover, the algorithms suggest the best way to alter the network to regain feasibility by providing realtime elastic resource management. Specifically, two approaches are proposed and evaluated. The first approach is the Irreducible Infeasible Set (IIS) Repair, and the second is the Minimum Cost Redesign. We evaluate the proposed algorithms using two practical use cases: an offline use case, where we need to fix the infeasibility for a bulk of service instances, and an online use case, where we need to fix the infeasibility for a single service instance in realtime. Furthermore, theoretical analysis and results show that the Minimum Cost Redesign method outperforms the IIS Repair method. Results also verify that our algorithms can provide practical solutions for both use cases in realtime.

Intent Based Networking (IBN) is a new paradigm that promises to create autonomous networks that can comply to high-level intents of network users and exhibit self-adaptation and self-optimization properties. The major IBN component to achieve this is network assurance. The assurance has as a goal to autonomously trigger corrective actions, whenever the conditions of the network do not allow to fulfill the performance requirements of its users. Accordingly, network assurance is expected to be largely based on monitoring and telemetry processes running conjointly with state of the art Artificial Intelligence techniques that will set off remedy solutions to bring the network into a compliance state. Given the importance of IBN and network assurance into the designing of autonomous networks, this paper introduces the main architectural components and technologies needed to achieve this visionary evolution of next generation networks. Particular emphasis is placed on how the network can interact with the end users and network operators in an easy way as mandated by the IBN premises. Finally, light is shed on the open challenges and future directions of this novel but unexplored research topic.

Recent advances in cloud computing and data centers have increased the demands for monitoring the network infrastructure and the applications that it hosts. The monitoring processes let network administrators to be aware of the status of the physical and logical units that compose their system. Since the goal of next generation networks is to minimise the administrators’ intervention, the alerting systems should minimize the frequency of notifications, emphasizing on critical scenarios such as when a monitoring metric surpasses a threshold or an anomalous behaviour is detected. However, current monitoring tools flood network administrators with hundreds of notifications every day. In this paper, we propose a binary classification approach, in order to decide if the administrators should be notified through monitoring alerts or not. To do so, our framework is build upon real monitoring logs and alerts, that show how the administrators reacted when receiving an alert. Extensive simulation results assess the performance of various classification approaches and reveal that random forests are great candidates for the binary classification alerting system that we propose, in terms of classification efficiency and computational overhead.

Recent trends in Network Function Virtualization (NFV) combined with Internet of Things (IoT) and 5G applications have reshaped the network service offering. In particular, Service Function Chains (SFCs) can associate network functions with physical and virtual resources towards providing a complete network service. Concurrently, the management of a continuously expanding network and the fulfillment of the applications’ requirements pave the way for autonomic network solutions. Intent Based Networking (IBN) is a novel paradigm that aims to achieve the automatic orchestration of network services and the assurance of their performance. Accordingly, in this paper, we propose a novel automated network assurance model, based on Model Predictive Control, to guarantee the Quality of Service (QoS) and security requirements of multi-tenant and IBN-enabled SFCs. In this context, corrective decisions are proactively taken, in the form of incoming intent relocations among the SFCs. The results reveal that our model can assure with high probability the application requirements and minimize QoS violations.

Modern networking paradigms like Service Function Chaining (SFC) allow for services to be broken down to a series of ordered and interconnected Virtualized Network Functions (VNFs) that can be hosted in generic servers in EdgeCloud datacenters. Nonetheless, a critical issue arises, when a hardware or software failure occurs and the VNFs of an SFC need to be repositioned, allowing to autonomously bring the system back to its normal operation, a process called self-healing. In this paper, a distributed methodology is proposed that aims to address this challenge, considering the requirements of all involved actors. Specifically, a Reinforcement Learning (RL) based algorithm is proposed that allows to iteratively optimize and determine an SFC healing solution upon a datacenter failure. As a second stage, a revenue-driven resource allocation mechanism is integrated, to resolve the contention for resources in an already functional datacenter that potentially occurs due to the repositioning. Various simulation scenarios prove the efficiency of our proposed resilient healing mechanism.

Cloud Computing has revolutionized the information technology world and the application offering over the last two decades. At the same time recent trends in Network Function Virtualization (NFV) and Software-Defined Wide Area Networks (SD-WAN) and the combination of those with the Cloud paradigm has allowed an unprecedented shift of enterprise networking services towards the Public Cloud. Even though this network evolutionary approach brings many benefits, it still presents many drawbacks as well. The performance stability and service continuity over a black box Public Cloud infrastructure can hinder the formal service guarantees that many new emerging applications may have. To this end, in this paper, we aim to shed light on the overall performance achieved when deploying coast-to-coast and intercontinental Service Function Chains (SFCs) that interconnect geographically distributed enterprise branches over the Amazon Web Services (AWS) infrastructure. In particular, we investigate the impact of region, Virtual Machine (VM) instance, time of the day and day of the week in the overall throughput and delay attained. The obtained results show the strengths and weaknesses of entirely relying on the AWS infrastructure to offer networking services by investigating possible hidden performance bottlenecks.

Internet of Things (IoT) along with Virtualized Network Function (VNFs) are creating a wide variety of opportunities for emerging vertical applications. Network operators are faced with a strategic puzzle on how to balance limited resource availability, dynamic IoT traffic requirements, and dynamic IoT device behavior in an end-to-end communication paradigm. To this end, this paper formally defines the IoT-aware VNF Placement (IVP) problem. We then evaluate an indicative set of placement algorithms with different objective functions under static and dynamic traffic scenarios to study their impact on the overall performance. The algorithms are evaluated based on realistic IoT traffic statistics in a smart-city environment and are presented as a simulation-based case study. Evaluation results emphasize the critical impact of considering multi-objective algorithms to accurately capture a set of conflicting goals, while efficiently balancing between them when solving the IVP problem. Finally, we shed light on the importance of using sophisticated lightweight approximation algorithms, to alleviate the inadequacies of the optimal mathematical solution.

Internet of Things (IoT) devices generate a tremendous amount of time series data that is extremely dynamic, heterogeneous and time dependent. Such types of data introduce significant challenges for the real-time prediction of QoS metrics of IoT applications with different traffic characteristics. To this end, in this paper, we propose a temporal transformer model and a unified system to predict several QoS metrics of heterogeneous IoT applications when they communicate with the Edge of the network. The transformer model also leverages an attention module to provide a solution for both short-term and long-term sequence prediction of QoS metrics that allows to better extract any time dependencies. In particular, in our framework, we firstly generate a set of datasets containing real-time traffic information of five different IoT applications such as Heating, Ventilation, and Air Conditioning (HVAC), lighting, Voice over Internet Protocol (VoIP), surveillance and emergency response using the 802.15.4 access technology and the RPL routing protocol. Following, we perform the data cleaning, downsampling and pre-processing of the datasets and we construct the QoS datasets, which include four QoS metrics, namely throughput, packet delivery ratio, packet loss ratio and latency. Finally, we evaluate the transformer model through extensive experimentation using both short-term and long-term dependencies and we show that our model can guarantee a robust performance and accurate QoS prediction.

Intent-Based Networking (IBN) is a recently proposed networking solution that allows networks to be configured and adapted autonomously according to the users' or operators' high-level intentions. However, a significant component of IBN is to assure that the network accurately and automatically deploys the intent throughout its lifecycle. To this end, in this study, we propose a network assurance solution for data center IBN networks. For the assurance model, we propose some specific data preparation procedures and Machine Learning (ML) models for the problem of time series forecasting. Specifically, we construct three main ML models that are based on the architecture of Convolutional Neural Networks (CNN) and Recurrent Neural Network (RNN). Our evaluation experiments, based on real data center Virtual Machine (VM) data traces, reveal the effectiveness of our methods in terms of CPU percentage usage prediction accuracy and speed. At the same time, our best-performing model can predict sufficiently far into the future with good accuracy.

Orchestration mechanisms play a pivotal role in assisting service providers in deploying their increasingly complex virtual network services seamlessly thanks to Network Function Virtualization (NFV) and Software-defined networking (SDN) technology enablers. Unfortunately, existing state-of-the-art orchestration techniques suffer from non-scalability and time-efficiency aptitude when the services require VNFs to be distributed across cloud and edge environments with complex dimensions. Furthermore, they provide competitive solutions in good execution time only for small-scale scenarios (e.g., in seconds for 50 nodes) but generally require an exorbitant amount of time to converge towards feasible solutions for medium-scale or even large-scale schemes. This paper proposes VAPNIC: an innovative approach that solves the VNF placement and chaining problem with lower algorithmic complexity using a disjoint-set data structure aided divide-and-conquer strategy. Our method's unique design is the non-use of any existing shortest path search algorithms to chain the virtual network functions. To the best of our knowledge, this is the first work that strives to tackle the placement and chaining of the VNFs from a distinctive perspective in the case of medium-and-large scale scenarios with a fast and scalable heuristic that exploits the divide-and-conquer design paradigm based on multi-branched recursion. Experimental results indicate that VAPNIC outperforms existing approaches in acceptance rate, resource utilization, scalability, and time efficiency.

IoT applications are quickly evolving in scope and objectives while their focus is being shifted toward supporting dynamic users’ requirements. IoT users initiate applications and expect quick and reliable deployment without worrying about the underlying complexities of the required sensing and routing resources. On the other hand, IoT sensing nodes, sinks, and gateways are heterogeneous, have limited resources, and require significant cost and installation time. Sensing network-level virtualization through virtual Sensing Networks (VSNs) could play an important role in enabling the formation of virtual groups that link the needed IoT sensing and routing resources. These VSNs can be initiated on-demand with the goal to satisfy different IoT applications’ requirements. In this context, we present a joint algorithm for IoT Sensing Resource Allocation with Dynamic Resource-Based Routing (SRADRR). The SRADRR algorithm builds on the current distinguished empowerment of sensing networks using recent standards like RPL and 6LowPAN. The proposed algorithm suggests employing the RPL standard concepts to create DODAG routing trees that dynamically adapt according to the available sensing resources and the requirements of the running and arriving applications. Our results and implementation of the SRADRR reveal promising enhancements in the overall applications deployment rate.

The proliferation of Internet of Things (IoT) and edge devices constitute important an efficient orchestration of the edge computing infrastructures, calling the providers to rethink their decision making methods. The resource usage prediction can be a prominent source of information for adaptive resource allocation and task offloading. In this research, we propose a Gated Recurrent Neural Network multi-output regression model that leverage time series resource usage metrics. The edge computing infrastructures are characterized as dynamical and heterogeneous environments. This motivated us to propose the innovative Hybrid Bayesian Evolutionary Strategy (HBES) algorithm for automated adaptation of the resource usage models in order to to enhance the generality of our approach. The proposed resource usage prediction mechanism has been experimentally evaluated and compared with other state of the art methods with significant improvements in terms of RMSE and MAE.

Network Function Virtualization (NFV) along with Service Function Chaining (SFC) has proliferated the way that network functions and services are introduced, offered, and deployed. In particular, NFV and SFC are the basic components of the so called network softwarization era, promoting significant cost reductions and agility. Nonetheless, a constant debate exist regarding the network performance achieved and the holistic establishment of NFV as the key and sole component of the network infrastructure. To this end, this paper tries to shed light in remedy solutions when chaining Virtualized Network Functions (VNFs) internally on a standard x86 server using the Single-Root Input Output Virtualization (SR-IOV) technology, in order to maximize the throughput and delay achieved.

The virtualization of network functions is promising significant cost reductions for network operators. Running multiple network functions on a standard x86 server instead of dedicated appliances can increase the utilization of the underlying hardware,while reducing the maintenance and management costs of such functions. However, total cost of ownership calculations are typically a function of the attainable network throughput, which in a virtualized system is highly dependent on the overall system architecture - in particular the input/output (I/O) path. In this paper we investigate the attainable performance of an x86 host running multiple virtualized network functions (VNFs) under different I/O architectures: OVS-DPDK, SR-IOV, and FD.io VPP. Running multiple VNFs in parallel on a standard x86 host is a common use-case for cloud-based networking services. We show that the system throughput in a multi-VNF environment differs significantly from deployments where only a single VNF is running on a server.

Network Function Virtualization promises to reduce the overall operational and capital expenses experienced by the network operators. Running multiple network functions on top of a standard x86 server instead of dedicated appliances can increase the utilization of the underlying hardware and reduce the maintenance and management costs. However, total cost of ownership calculations are typically a function of the attainable network throughput, which in a virtualized system is highly dependent on the overall system architecture - in particular the input/ output (I/O) path. In this paper, we investigate the attainable performance of an x86 host running multiple Virtualized Network Functions (VNFs) under different I/O architectures: OVS, SRIOV and FD.io VPP. We show that the system throughput in a multi-VNF environment differs significantly from deployments where only a single VNF is running on a server, while different I/O architectures can achieve different levels of performance.

Network function virtualization along with network service chaining and forwarding graphs envision a reduction in the respective cost that end users, service providers, and network operators are experiencing, while providing complete and high quality services. The allocation of these service chains in a pool of available cloud or data center resources is a challenging problem that can affect the overall performance of the offered network services. Furthermore, a number of challenges associated with the hardware capabilities and the available resources of the cloud infrastructure, along with possible collocation constraints between the components of the service chain, can exponentially increase the complexity of resource allocation. This paper examines how to improve the overall allocation performance of deploying service chains in a cloud environment satisfying server affinity, collocation, and latency constraints. The proposed method is inspired by a partitioning game, where the various components of a service chain are split in a set of partitions executed as virtual machines/containers in appropriate servers. We mathematically prove that a Nash equilibrium exists for our partitioning game corresponding to an optimal solution. By implementing the partitioning game as an iterative refinement process, we also experimentally validate that the proposed algorithm converges to the optimal solution.

Network Function Virtualization along with Network Service Chaining envision a reduction in the respective cost that end users, service providers, and network operators are experiencing, while providing complete and high quality services. However, the vast range of available services and the service on-demand model, creates dynamic traffic conditions that necessitates a flexible and automatic network platform to redirect traffic according to network conditions. In this paper, we study the problem of deploying service chains, consisting of a number of virtualized network functions (VNFs), in a SDN enabled data center network, where a random number of users are associated with each service chain. To this end, appropriate resource management algorithms are introduced for the placement of VNFs satisfying server affinity and latency constraints. The interconnection of the VNFs is facilitated by an SDN controller, which periodically recalculates the routing paths to adjust to the dynamic traffic conditions.

Cloud computing has emerged as the computing paradigm that enables the delivery of utility-based IT services to users. The hyper-growth of Cloud computing has led to increased power consumption with significant consequences both in terms of environmental and operational costs. Hence, over the last years, attention has been drawn to optimizing energy consumption at the data center, aimed at the reduction of carbon footprints. However the world of Cloud Computing is constantly developing, with new concepts introduced while additional challenges arise. In this paper, a method for energy efficient resource allocation is proposed, in the context of a networked cloud environment. The method employs dynamic server consolidation by periodic VM migration. The approach is validated conducting performance evaluation via simulation, while it is compared against energy aware / non energy aware methods, using a set of both power indication and resource allocation metrics.

Cloud computing builds upon advances on virtualization and distributed computing to support cost-efficient usage of computing resources, emphasizing on resource scalability and on demand services. Moving away from traditional data-center oriented models, distributed clouds extend over a loosely coupled federated substrate, offering enhanced communication and computational services to target end-users with quality of service (QoS) requirements, as dictated by the future Internet vision. Toward facilitating the efficient realization of such networked computing environments, computing and networking resources need to be jointly treated and optimized. This requires delivery of user-driven sets of virtual resources, dynamically allocated to actual substrate resources within networked clouds, creating the need to revisit resource mapping algorithms and tailor them to a composite virtual resource mapping problem. In this paper, toward providing a unified resource allocation framework for networked clouds, we first formulate the optimal networked cloud mapping problem as a mixed integer programming (MIP) problem, indicating objectives related to cost efficiency of the resource mapping procedure, while abiding by user requests for QoS-aware virtual resources. We subsequently propose a method for the efficient mapping of resource requests onto a shared substrate interconnecting various islands of computing resources, and adopt a heuristic methodology to address the problem. The efficiency of the proposed approach is illustrated in a simulation/emulation environment, that allows for a flexible, structured, and comparative performance evaluation. We conclude by outlining a proof-of-concept realization of our proposed schema, mounted over the European future Internet test-bed FEDERICA, a resource virtualization platform augmented with network and computing facilities.

Cloud-oriented content delivery networks (CCDNs) constitute a promising alternative to traditional content delivery networks. Exploiting the advantages and principles of the cloud, such as the pay as you go business model and geographical dispersion of resources, CCDN can provide a viable and cost-effective solution for realizing content delivery networks and services. In this paper, a hierarchical framework is proposed and evaluated toward an efficient and scalable solution of content distribution over a multiprovider networked cloud environment, where inter and intra cloud communication resources are simultaneously considered along with traditional cloud computing resources. To efficiently deal with the CCDN deployment problem in this emerging and challenging computing paradigm, the problem is decomposed to graph partitioning and replica placement problems while appropriate cost models are introduced/adapted. Novel approaches on the replica placement problem within the cloud are proposed while the limitations of the physical substrate are taken into consideration. The performance of the proposed hierarchical CCDN framework is assessed via modeling and simulation, while appropriate metrics are defined/adopted associated with and reflecting the interests of the different identified involved key players.

The cloud represents a computing paradigm where shared configurable resources are provided as a service over the Internet. Adding intra- or intercloud communication resources to the resource mix leads to a networked cloud computing environment. Following the cloud infrastructure as a Service paradigm and in order to create a flexible management framework, it is of paramount importance to address efficiently the resource mapping problem within this context. To deal with the inherent complexity and scalability issue of the resource mapping problem across different administrative domains, in this paper a hierarchical framework is described. First, a novel request partitioning approach based on Iterated Local Search is introduced that facilitates the cost-efficient and online splitting of user requests among eligible cloud service providers (CPs) within a networked cloud environment. Following and capitalizing on the outcome of the request partitioning phase, the embedding phase-where the actual mapping of requested virtual to physical resources is performed can be realized through the use of a distributed intracloud resource mapping approach that allows for efficient and balanced allocation of cloud resources. Finally, a thorough evaluation of the proposed overall framework on a simulated networked cloud environment is provided and critically compared against an exact request partitioning solution as well as another common intradomain virtual resource embedding solution.