Publications
All our publications can be found in 
Beyond Multi-Access Edge Computing: Essentials to Realize a Mobile, Constrained Edge
The main purpose of ETSI multi-access edge computing (MEC) is to improve latency and bandwidth consumption by keeping local traffic local while providing computing resources near the end-user. Despite its clear benefits, the next-generation of hyper-distributed applications (e.g., edge robotics, augmented environments, or smart agriculture) will exacerbate latency and bandwidth requirements, posing significant challenges to today’s MEC deployments.
A Survey on IoT-Edge-Cloud Continuum Systems: Status, Challenges, Use Cases, and Open Issues
The rapid growth in the number of interconnected devices on the Internet (referred to as the Internet of Things—IoT), along with the huge volume of data that are exchanged and processed, has created a new landscape in network design and operation. Due to the limited battery size and computational capabilities of IoT nodes, data processing usually takes place on external devices. Since latency minimization is a key concept in modern-era networks, edge servers that are in close proximity to IoT nodes gather and process related data, while in some cases data offloading in the cloud might have to take place.
Distributed Machine Learning and Native AI Enablers for End-to-End Resources Management in 6G
6G targets a broad and ambitious range of networking scenarios with stringent and diverse requirements. Such challenging demands require a multitude of computational and communication resources and means for their efficient and coordinated management in an end-to-end fashion across various domains. Conventional approaches cannot handle the complexity, dynamicity, and end-to-end scope of the problem, and solutions based on artificial intelligence (AI) become necessary.
Learning to Fulfill the User Demands in 5G-enabled Wireless Networks through Power Allocation: a Reinforcement Learning approach
The goal of the study presented in this paper is to evaluate the performance of a proposed Reinforcement Learning (RL) power allocation algorithm. The algorithm follows a demand-driven power adjustment approach aiming at maximizing the number of users inside a coverage area that experience the requested throughput to accommodate their needs. In this context, different Quality of Service (QoS) classes, corresponding to different throughput demands, have been taken into account in various simulation scenarios.
Uplink NOMA for UAV-Aided Maritime Internet-of-Things
Maritime activities are vital for economic growth, being further accelerated by various emerging maritime Internet of Things (IoT) use cases, including smart ports, autonomous navigation, and ocean monitoring systems. However, broadband, low-delay, and reliable wireless connectivity to the ever-increasing number of vessels, buoys, platforms and sensors in maritime communication networks (MCNs) has not yet been achieved. Towards this end, the integration of unmanned aerial vehicles (UAVs) in MCNs provides an aerial dimension to current deployments, relying on shore-based base stations (BSs) with limited coverage and satellite links with high latency.
A Stacking Ensemble Learning Model for Waste Prediction in Offset Printing
The production of quality printing products requires a highly complex and uncertain process, which leads to the unavoidable generation of printing defects. This common phenomenon has severe impacts on many levels for Offset Printing manufacturers, ranging from a direct economic loss to the environmental impact of wasted resources. Therefore, the accurate estimation of the amount of paper waste expected during each press run, will minimize the paper consumption while promoting environmentally sustainable principles.
Data Aging Matters: Federated Learning-Based Consumption Prediction in Smart Homes via Age-Based Model Weighting
Smart homes, powered mostly by Internet of Things (IoT) devices, have become very popular nowadays due to their ability to provide a holistic approach towards effective energy management. This is made feasible via the deployment of multiple sensors, which enables predicting energy consumption via machine learning approaches. In this work, we propose FedTime, a novel federated learning approach for predicting smart home consumption which takes into consideration the age of the time series datasets of each client.
Presentation Open Access SWForum.eu The Way Forward: Workshop on Future Challenges in Software Engineering
SWForum.eu Way Forward Workshop: Future Challenges in Software Engineering, held on 27 June 2023 at the Politecnico di Milano (POLIMI), in the Dipartamento di Elettronica, Informazione e Bioingegneria (DEIB), featured a dynamic agenda that encompassed a comprehensive range of topics and discussions.
Runtime security monitoring by an interplay between rule matching and deep learning-based anomaly detection on logs
In the era of digital transformation the increasing vulnerability of infrastructure and applications is often tied to the lack of technical capability and the improved intelligence of the attackers. In this paper, we discuss the complementarity between static security monitoring of rule matching and an application of self-supervised machine-learning to cybersecurity. Moreover, we analyse the context and challenges of supply chain resilience and smart logistics. Furthermore, we put this interplay between the two complementary methods in the context of a self-learning and self-healing approach.
Scaling Serverless Functions in Edge Networks: A Reinforcement Learning Approach
With rapid advances in containerization techniques, the serverless computing model is becoming a valid candidate execution model in edge networking, similar to the widely used cloud model for applications that are stateless, single purpose and event-driven, and in particular for delay-sensitive applications. One of the cloud serverless processes, i.e., the auto-scaling mechanism, cannot be however directly applied at the edge, due to the distributed nature of edge nodes, the difficulty of optimal resource allocation, and the delay sensitivity of workloads. We propose a solution to the auto-scaling problem by applying reinforcement learning (RL) approach to solving problem of efficient scaling and resource allocation of serverless functions in edge networks. We compare RL and Deep RL algorithms with empirical, monitoring-based heuristics, considering delay-sensitive applications. The simulation results shows that RL al-gorithm outperforms the standard, monitoring-based algorithms in terms of total delay of function requests, while achieving an improvement in delay performance by up to 50%.
Federated Learning-Aided Prognostics in the Shipping 4.0: Principles, Workflow, and Use Cases
In this work, we explore the integration of FL into PdM to support Shipping 4.0 applications, by using real datasets from the maritime sector. More specifically, we present the main FL principles, the proposed workflow and then, we evaluate and compare various FL algorithms in three maritime use cases, i.e. regression to predict the naval propulsion gas turbine (GT) measures, classification to predict the ship engine condition, and time-series regression to predict ship fuel consumption. The efficiency of the proposed FL-based PdM highlights its ability to improve maintenance decision-making, reduce downtime in the shipping industry, and enhance the operational efficiency of shipping fleets. The findings of this study support the advancement of PdM methodologies in Shipping 4.0, providing valuable insights for maritime stakeholders to adopt FL, as a viable and privacy-preserving solution, facilitating model sharing in the shipping industry and fostering collaboration opportunities among them.
Towards Optimal Serverless Function Scaling in Edge Computing Network
Serverless computing has emerged as a new execution model which gained a lot of attention in cloud computing thanks to the latest advances in containerization technologies. Recently, serverless has been adopted at the edge, where it can help overcome heterogeneity issues, constrained nature and dynamicity of edge devices. Due to the distributed nature of edge devices, however, the scaling of serverless functions presents a major challenge. We address this challenge by studying the optimality of serverless function scaling. To this end, we propose Semi-Markov Decision Process-based (SMDP) theoretical model, which yields optimal solutions by solving the serverless function scaling problem as a decision making problem. We compare the SMDP solution with practical, monitoring-based heuristics. We show that SMDP can be effectively used in edge computing networks, and in combination with monitoring-based approaches also in real-world implementations.
A Dataflow-Oriented Approach for Machine-Learning-Powered Internet of Things Applications
The rise of the Internet of Things (IoT) has led to an exponential increase in data generated by connected devices. Machine Learning (ML) has emerged as a powerful tool to analyze these data and enable intelligent IoT applications. However, developing and managing ML applications in the decentralized Cloud-to-Things continuum is extremely complex. This paper proposes Zenoh-Flow, a dataflow programming framework that supports the implementation of End-to-End (E2E) ML pipelines in a fully decentralized manner and abstracted from communication aspects. Thus, it simplifies the development and upgrade process of the next-generation ML-powered applications in the IoT domain. The proposed framework was demonstrated using a real-world use case, and the results showcased a significant improvement in overall performance and network usage compared to the original implementation. Additionally, other of its inherent benefits are a significant step towards developing efficient and scalable ML applications in the decentralized IoT ecosystem.
Zenoh-Unifying communication, storage, and computation from the cloud to the microcontroller
An increasing number of systems span from the data-center down to the micro-controller and need to smoothly operate across this continuum composed by extremely heterogeneous network technologies and computing platforms. Building these systems is quite challenging due to limitations of existing technological stacks. This paper introduces Zenoh a Pub/Sub- /Query protocol that unifies data at rest, data in motion and computations. Zenoh has been designed ground-up to address the needs of the cloud to micro-controller continuum. It has a minimal wire overhead of 5 bytes, it runs and perform on constrained as well as on high end networks and hardware.
New challenges in the implementation of digital projects in agriculture
Presentation about new challenges in the implementation of digital projects in agriculture.
Potential of Implementing Operational Metasystems in Agriculture Using the ICOS Project as an Example
Presentation about the potential of implementing operational metasystems in agriculture using the ICOS project as an example.
Hierarchical Management of extreme-scale task-based applications
The scale and heterogeneity of exascale systems increment the complexity of programming applications exploiting them. Task-based approaches with support for nested tasks are a good-fitting model for them because of the flexibility lying in the task concept. Resembling the hierarchical organization of the hardware, this paper proposes establishing a hierarchy in the application workflow for mapping coarse-grain tasks to the broader hardware components and finer-grain tasks to the lowest levels of the resource hierarchy to benefit from lower-latency and higher-bandwidth communications and exploiting locality. Building on a proposed mechanism to encapsulate within the task the management of its finer-grain parallelism, the paper presents a hierarchical peer-to-peer engine orchestrating the execution of workflow hierarchies with fully-decentralized management. The tests conducted on the MareNostrum 4 supercomputer using a prototype implementation prove the validity of the proposal supporting the execution of up to 707,653 tasks using 2,400 cores and achieving speedups of up to 106 times faster than executions of a single workflow and centralized management.
Enhanced Smart Advertising through Federated Learning
Smart advertising is growing in popularity and affecting businesses. Smart advertising is a more friendly, interactive, personalised, and creative method of promoting a product, and attempts to delight clients. AROUND is a social networking service that emphasizes smart advertising through an effective recommender system. The system considers user profiles, history, social network connections, mood, and IoT-supported positioning to select the most relevant ads using machine learning technology. Although the current deployment of the AROUND system is based on the cloud, an edge-based architecture provides relevant improvement in terms of system response time. In this paper we extend the edge-based strategy to leverage the potential of federated learning on multiple distributed edge servers. We show that federated learning can take advantage of the distributed nature of the system, and leverage the specificities of local features. In fact, in this research, we propose a novel federated learning solution to provide smart advertising as a classification problem which uses ensemble methods and logistic regression as internal (local) models and meta-heuristic algorithms for federated learning aggregation. As part of the experiments, we prove this technology on a real data set with more than one million registers, and show the efficiency in terms of enhanced accuracy and improved training and response speed.
Development of the EDGE-CLOUD solutions across domain
Edge-to-Cloud is expected to provide the means for workloads execution and data processing both in the Edge and the Cloud. Within this presentation we present different efforts addressing the challenges towards achieving the next generation of continuum management. We will explain this presenting efforts of two projects illuMINEation and ICOS. ICOS is proposing a high-level meta operating system (metaOS) to realize the continuum. The use case that is targeting to exploit the technologies is related to agriculture and robotics - the Agriculture Operational Robotic Platform (AORP) is an agro robot that can execute different tasks and missions, like sowing and tending crops, removing weeds, monitoring crop development, and identifying threats. The platform moves autonomously through the field, performing the assigned missions. The robotic platform consists of control and driving modules. In addition, it is equipped with interchangeable tools - a seeder and a sprayer. The AORP is equipped with cameras, sensors and Edge computational devices that can be connected to the Cloud directly, via the transport platform, or via farm connectivity.
A comprehensive survey on reinforcement-learning-based computation offloading techniques in Edge Computing Systems
This paper presents a comprehensive and detailed survey, where we analyse and classify the research papers in terms of use cases, network and edge computing architectures, objectives, RL algorithms, decision-making approaches, and time-varying characteristics considered in some scenarios, particularly on those related to the computation offloading problem in edge systems.
An NIDS for Known and Zero-Day Anomalies
Rapid development in the network infrastructure has resulted in sophisticated attacks which are hard to detect using typical network intrusion detection systems (NIDS). There is a strong need for efficient NIDS to detect these known attacks along with ever-emerging zero-day exploits. Existing NIDS are more focused on detecting known attacks using supervised machine learning approaches, achieving better performance for known attacks but poor detection of unknown attacks. Many NIDS have utilized the unsupervised approach, which results in better detection of unknown anomalies. In this paper, we proposed a Hybrid NIDS based on Semisupervised One-Class Support Vector Machine (OC-SVM) and Supervised Random Forest (RF) algorithms. This detection system has several stages. The First stage is based on OC-SVM, which filters benign and malicious traffic. The next stages use many parallel supervised models and an additional OC-SVM model to separate known and unknown attacks from malicious traffic. The previous process is done so that known attacks are classified by their type, and unknown attacks are detected. The proposed NIDS is tested on the standard public dataset CSE-CIC-IDS-2018. The evaluation results show that the system achieves a high accuracy, 99.45%, for detecting known attacks. Our proposed NIDS achieves an accuracy of 93.99% for unknown or zero-day attacks. The overall accuracy of the proposed NIDS is 95.95%. The system significantly improves the detection of known and unknown anomalies using a hybrid approach.
Security in DevSecOps: Applying Tools and Machine Learning to Verification and Monitoring Steps
Security represents one of the crucial concerns when it comes to DevOps methodology-empowered software development and service delivery process. Considering the adoption of Infrastructure as Code (IaC), even minor flaws could potentially cause fatal consequences, especially in sensitive domains such as healthcare and maritime applications. However, most of the existing solutions tackle either Static Application Security Testing (SAST) or run-time behavior analysis distinctly. In this paper, we propose a) IaC Scan Runner, an open-source solution developed in Python for inspecting a variety of state-of-the-art IaC languages in application design time and b) the run time anomaly detection tool called LOMOS. Both tools work in synergy and provide a valuable contribution to a DevSecOps tool set. The proposed approach is demonstrated and their results will be demonstrated on various case studies showcasing the capabilities of static analysis tool IaC Scan Runner combined with LOMOS – log analysis artificial intelligence-enabled framework.
5G/6G Architecture Evolution for XR and Metaverse: Feasibility Study, Security, and Privacy Challenges for Smart Culture Applications
This paper investigates the evolution of 5G/6G architectures to support demanding Extended Reality (XR) and Metaverse applications, focusing specifically on the ‘‘smart culture’’ domain. We evaluate the capabilities of the 5G Service-Based Architecture (SBA), including Multi-Access Edge Computing (MEC) and network analytics, through a comprehensive feasibility study comparing stringent XR requirements (bitrate, latency, capacity, power, accuracy) against current 5G performance. Our key contribution is the identification of significant performance gaps where 5G struggles to meet the demands of advanced XR, particularly concerning capacity, scalability, and ultra-low latency. Furthermore, we provide a detailed analysis of critical security and privacy challenges inherent in 5G-enabled XR environments, including virtualization vulnerabilities, API security, and sensitive data protection. While 5G provides core capabilities, significant challenges persist, emphasizing the need for continued research and the evolution toward 6G to effectively support immersive experiences in smart culture and the Metaverse.
Data-centric Service-Based Architecture for Edge-Native 6G Network
Starting with 5G, mobile networks moved away from the point-to-point model used by previous generations towards a Service-Based Architecture (SBA) focused on a Cloud-native design. While 5G considers the SBA mainly as a central location with a single administrative domain, 6G is expected to become more of a fully distributed system. This raises the need for flexible service routing capabilities that allow the utilization of the distributed resources available over multiple infrastructures. This paper embraces this vision and explores the utilization of data-centric and dataflow mechanisms for a seamless realization and interaction of composable services in a fully distributed Edgenative 6G architecture. The feasibility and suitability of this vision are showcased through a proof-of-concept prototype, validated over selected 5G workflows using different technologies, with results demonstrating the advantages of pursuing data-centric and dataflow approaches.
HOODIE: Hybrid Computation Offloading via Distributed Deep Reinforcement Learning in Delay-Aware Cloud-Edge Continuum
Cloud-Edge Computing Continuum (CEC) system, where edge and cloud nodes are seamlessly connected, is dedicated to handle substantial computational loads offloaded by end-users. These tasks can suffer from delays or be dropped entirely when deadlines are missed, particularly under fluctuating network conditions and resource limitations. The CEC is coupled with the need for hybrid task offloading, where the task placement decisions concern whether the tasks are processed locally, offloaded vertically to the cloud, or horizontally to interconnected edge servers. In this paper, we present a distributed hybrid task offloading scheme (HOODIE) designed to jointly optimize the tasks latency and drop rate, under dynamic CEC traffic. HOODIE employs a model-free deep reinforcement learning (DRL) framework, where distributed DRL agents at each edge server autonomously determine offloading decisions without global task distribution awareness. To further enhance the system pro-activity and learning stability, we incorporate techniques such as Long Short-term Memory (LSTM), Dueling deep Q-networks (DQN), and double-DQN. Extensive simulation results demonstrate that HOODIE effectively reduces task drop rates and average task processing delays, outperforming several baseline methods under changing CEC settings and dynamic conditions.
System Level Performance Assessment of Large-Scale Cell-Free Massive MIMO Orientations With Cooperative Beamforming
The goal of the study presented in this paper is to evaluate the performance of a proposed adaptive beamforming approach in cell-free massive multiple input multiple output (CF-mMIMO) orientations. To this end, mobile stations (MSs) can be served by multiple access points (APs) simultaneously. In the same context, the performance of a dynamic physical resource block (PRB) allocation approach is evaluated as well, where the set of assigned PRBs per active MS is constantly updated according to their signal strength and the amount of interference that cause to the rest of the co-channel MSs. Performance evaluation takes place in a two-tier wireless orientation, employing a system-level simulator designed for parallel Monte Carlo simulations. According to the presented results, a significant gain in energy efficiency (EE) can be achieved for medium data rate services when comparing the cell-free (CF) resource allocation approach to single AP links (non-CF). This is made feasible via cooperative beamforming, where on one hand, the radiation figures of the APs that serve a particular MS are jointly updated to ensure quality of service (QoS), and on the other hand, the effects of these updates on the other MSs are evaluated as well. Although EE for high data rate services decreases compared to the non-CF scenario, the proposed dynamic PRB allocation strategy significantly lowers the number of active radiating elements required to meet minimum QoS standards, thereby reducing both hardware and computational demands.
Intelligent Edge Computing and Machine Learning: A Survey of Optimization and Applications
Intelligent edge machine learning has emerged as a paradigm for deploying smart applications across resource-constrained devices in next-generation network infrastructures. This survey addresses the critical challenges of implementing machine learning models on edge devices within distributed network environments, including computational limitations, memory constraints, and energy-efficiency requirements for real-time intelligent inference. We provide comprehensive analysis of soft computing optimization strategies essential for intelligent edge deployment, systematically examining model compression techniques including pruning, quantization methods, knowledge distillation, and low-rank decomposition approaches. The survey explores intelligent MLOps frameworks tailored for network edge environments, addressing continuous model adaptation, monitoring under data drift, and federated learning for distributed intelligence while preserving privacy in next-generation networks. Our work covers practical applications across intelligent smart agriculture, energy management, healthcare, and industrial monitoring within network infrastructures, highlighting domain-specific challenges and emerging solutions. We analyze specialized hardware architectures, cloud offloading strategies, and distributed learning approaches that enable intelligent edge computing in heterogeneous network environments. The survey identifies critical research gaps in multimodal model deployment, streaming learning under concept drift, and integration of soft computing techniques with intelligent edge orchestration frameworks for network applications. These gaps directly manifest as open challenges in balancing computational efficiency with model robustness due to limited multimodal optimization techniques, developing sustainable intelligent edge AI systems arising from inadequate streaming learning adaptation, and creating adaptive network applications for dynamic environments resulting from insufficient soft computing integration. This comprehensive roadmap synthesizes current intelligent edge machine learning solutions with emerging soft computing approaches, providing researchers and practitioners with insights for developing next-generation intelligent edge computing systems that leverage machine learning capabilities in distributed network infrastructures.
Offloading Artificial Intelligence Workloads across the Computing Continuum by means of Active Storage Systems
The increasing demand for artificial intelligence (AI) workloads across diverse computing environments has driven the need for more efficient data management strategies. Traditional cloud-based architectures struggle to handle the sheer volume and velocity of AI-driven data, leading to inefficiencies in storage, computation, and data movement. This paper explores the integration of active storage systems within the computing continuum to optimize AI workload distribution. By embedding computation directly into storage architectures, active storage is able to reduce data transfer overhead, enhancing performance and improving resource utilization. Other existing frameworks and architectures offer mechanisms to distribute certain AI processes across distributed environments; however, they lack the flexibility and adaptability that the continuum requires, both regarding the heterogeneity of devices and the rapid-changing algorithms and models being used by domain experts and researchers. This article proposes a software architecture aimed at seamlessly distributing AI workloads across the computing continuum, and presents its implementation using mainstream Python libraries and dataClay, an active storage platform. The evaluation shows the benefits and trade-offs regarding memory consumption, storage requirements, training times, and execution efficiency across different devices. Experimental results demonstrate that the process of offloading workloads through active storage significantly improves memory efficiency and training speeds while maintaining accuracy. Our findings highlight the potential of active storage to revolutionize AI workload management, making distributed AI deployments more scalable and resource-efficient with a very low entry barrier for domain experts and application developers.
SBNNR: Small-Size Bat-Optimized KNN Regression
Small datasets are frequent in some scientific fields. Such datasets are usually created due to the difficulty or cost of producing laboratory and experimental data. On the other hand, researchers are interested in using machine learning methods to analyze this scale of data. For this reason, in some cases, low-performance, overfitting models are developed for small-scale data. As a result, it appears necessary to develop methods for dealing with this type of data. In this research, we provide a new and innovative framework for regression problems with a small sample size. The base of our proposed method is the K-nearest neighbors (KNN) algorithm. For feature selection, instance selection, and hyperparameter tuning, we use the bat optimization algorithm (BA). Generative Adversarial Networks (GANs) are employed to generate synthetic data, effectively addressing the challenges associated with data sparsity. Concurrently, Deep Neural Networks (DNNs), as a deep learning approach, are utilized for feature extraction from both synthetic and real datasets. This hybrid framework integrates KNN, DNN, and GAN as foundational components and is optimized in multiple aspects (features, instances, and hyperparameters) using BA. The outcomes exhibit an enhancement of up to 5% in the coefficient of determination (R2 score) using the proposed method compared to the standard KNN method optimized through grid search.
Adaptive Machine Learning for Resource-Constrained Environments
The Internet of Things is an example domain where data is perpetually generated in ever-increasing quantities, reflecting the proliferation of connected devices and the formation of continuous data streams over time. Consequently, the demand for ad-hoc, cost-effective machine learning solutions must adapt to this evolving data influx. This study tackles the task of offloading in small gateways, exacerbated by their dynamic availability over time. An approach leveraging CPU utilization metrics using online and continual machine learning techniques is proposed to predict gateway availability. These methods are compared to popular machine learning algorithms and a recent time-series foundation model, Lag-Llama, for fine-tuned and zero-shot setups. Their performance is benchmarked on a dataset of CPU utilization measurements over time from an IoT gateway and focuses on model metrics such as prediction errors, training and inference times, and memory consumption. Our primary objective is to study new efficient ways to predict CPU performance in IoT environments. Across various scenarios, our findings highlight that ensemble and online methods offer promising results for this task in terms of accuracy while maintaining a low resource footprint. Code is available at https://github.com/sebasmos/AML4CPU
An Edge/Cloud Continuum with Wearable Kinetic Energy Harvesting IoT Devices in Remote Areas
One of the key factors critical to the advancements of IoT systems in remote areas today are energy-efficient IoT deployment and the integration with IoT/edge/continuum. An energy-efficient IoT deployment requires finding adequate solutions for applications that require remote area devices and the related replacement and charging of batteries. On the other hand, an efficient integration of different communication technologies spanning the IoT, edge and cloud continuum that at the same time can integrate energy harvesting devices in remote areas is still an open challenge. In this paper, we integrate energy harvesting with wearable remote IoT devices on freely roaming farm animals within the edge/cloud continuum along its powerful application layer protocols, MQTT and AMQP. We experimentally investigate the performance of kinetic energy harvester used to power a LoRa module to send application layer messages from IoT to cloud. From the functional system testing perspective, we show that these messages can be successfully forwarded for further processing and evaluation in the edge and cloud setting even from the remote areas. We engineered an inexpensive and first open-source multi-protocol MQTT based communication gateway, as an alternative to today’s proprietary and expensive gateway solutions, and we built a system that can not only power the capturing of animal movement patterns outdoors, but also the related application-layer protocol messages.
An Experimental Study of the Response Time in an Edge-Cloud Continuum with ClusterLink
In this paper, we conduct an experimental study to provide a general sense of the application response time implications that inter-cluster communication experiences at the edge at the example of a specific IoT-edge-cloud contiuum solution from the EU Project ICOS called ClusterLink. We create an environment to emulate different networking topologies that include multiple cloud or edge sites scenarios, and conduct a set of tests to compare the application response times via ClusterLink to direct communications in relation to node distances and request/response payload size. Our results show that, in an edge context, ClusterLink does not introduce a significant processing overhead to the communication for small payloads as compared to cloud. For higher payloads and on comparably more aged consumer hardware, ClusterLink version 0.2 introduces communication overhead relative to the delay experienced on the link.
Benchmarking Performance of Various MQTT Broker Implementations in a Compute Continuum
With the increasing adoption of IoT devices and applications, significant research and development efforts have been centered around engineering novel ecosystems referred to as the IoT-edge-cloud compute continuum. In this article, we implement, analyze and present a case study for performance benchmarking of five well known and select open source MQTT broker implementations in an open-source compute continuum testbed. The proposed MQTT broker implementations are evaluated in terms of response time, different payload sizes and throughput. Measurements and results show that the hardware platform used, the message size, as well as the network parameters (latency, packet loss and jitter) have a significant impact on the resulting performance of various broker implementations and therefore have to be carefully considered in the selection process for the building blocks of the continuum. All implementations and measurements are made to be fully reproducible and free and open source.
COOLER: Cooperative Computation Offloading in Edge-Cloud Continuum Under Latency Constraints via Multi-Agent Deep Reinforcement Learning
In the burgeoning domain of the edge-cloud con-tinuum (ECC), the efficient management of computational tasks offloaded from mobile devices to edge nodes is paramount. This paper introduces a Cooperative cOmputation Offloading scheme for ECC via Latency-aware multi-agent Reinforcement learning (COOLER), a distributed framework designed to address the challenges posed by the uncertain load dynamics at edge nodes. COOLER enables each edge node to autonomously make offloading decisions, optimizing for non-divisible, delay-sensitive tasks without prior knowledge of other nodes‘ task models and decisions. By formulating a multi-agent computation offloading problem, COOLER aims to minimize the expected long-term latency and task drop ratio. Following the ECC requirements for seamless task flow both within Edge layer and between Edge-Cloud layers, COOLER considers that task computation decisions are three-fold: (i) local computation, (ii) horizontal offloading to another edge node, or (iii) vertical offloading to the Cloud. The integration of advanced techniques such as long short-term memory (LSTM), double deep Q-network (DQN) and dueling DQN enhances the estimation of long-term costs, thereby improving decision-making efficacy. Simulation results demonstrate that COOLER significantly outperforms baseline offloading algorithms, reducing both the ratio of dropped tasks and average delay, and better harnessing the processing capacities of edge nodes.
ClusterLink: Redefining Application Connectivity for the Multi-cloud Era
Modern software development abstracts applications from the underlying infrastructure, enabling global-scale deployment with minimal concern about low-level networking details. However, when these infrastructure-agnostic software components need to communicate, they encounter significant networking limitations. This forces developers to either navigate complex, low-level networking constructs to achieve the desired connectivity or give up on truly flexible connectivity and limit their software to static connectivity patterns. In this paper, we focus on the evolving challenges of application connectivity in today’s hyper-distributed reality. We propose to model connectivity around the notion of application services and have realized this proposal as ClusterLink, which exposes the app-level APIs for specifying communication policies at a very granular level and implements them efficiently. This paper shares details on ClusterLink design principles, APIs, architecture, and implementation, and shows that ClusterLink outperforms its closest competitor by 2.5x in throughput in a cloud-based experimental setting.
Deploying AI-Based Applications with Serverless Computing in 6G Networks: An Experimental Study
Future 6G networks are expected to heavily utilize machine learning capabilities in a wide variety of applications with features and benefits for both, the end user and the provider. While the options for utilizing these technologies are almost endless, from the perspective of network architecture and standardized service, the deployment decisions on where to execute the AI-tasks are critical, especially when considering the dynamic and heterogeneous nature of processing and connectivity capability of 6G networks. On the other hand, conceptual and standardization work is still in its infancy, as to how to categorize ML applications in 6G landscapes; some of them are part of network management functions, some target the inference itself, while many others emphasize model training. It is likely that future mobile services may all be in the AI domain, or combined with AI. This work makes a case for the serverless computing paradigm to be used to this end. We first provide an overview of different machine learning applications that are expected to be relevant in 6G networks. We then create a set of general requirements for software engineering solutions executing these workloads from them and propose and implement a high-level edge-focused architecture to execute such tasks. We then map the ML-serverless paradigm to the case study of 6G architecture and test the resulting performance experimentally for a machine learning application against a setup created in a more traditional, cloud-based manner. Our results show that, while there is a tradeoff in predictability of the response times and the accuracy, the achieved median accuracy in a 6G setup remains the same, while the median response time decreases by around 25% compared to the cloud setup.
Evaluating the Impact of Inter-Cluster Communications in Edge Computing
Distributed applications based on micro-services in edge computing are becoming increasingly popular. Kubernetes is the default framework for orchestrating and managing micro-service-based applications. Notwithstanding, the requirement to run applications between multiple sites at cloud and edge poses new challenges since Kubernetes does not natively provide tools to abstract inter-cluster communications at the application level. In this paper, we evaluate for the first time the impact of inter-cluster communication on edge computing performance by using three prominent, open-source inter-cluster communication projects and tools (Submariner, ClusterLink, and Skupper). We develop a fully open-source testbed that integrates these tools modularly and experimentally benchmark sample applications on their performance running in a multi-cluster edge computing system under varying networking conditions. We experimentally analyze two classes of envisioned mobile applications (industrial automation and vehicle decision drive assist). Our results show that ClusterLink performs best out of the three tools in scenarios with increased payloads regardless of the underlying networking conditions or transmission direction. Skupper closely follows it unless requests and replies transport big payloads. Finally, for small payloads, Submariner slightly outperforms the other tools.
ICOS: An Intelligent MetaOS for the Continuum
This paper presents ICOS, a meta-operating system designed for the cloud continuum. The paper provides insight into the ICOS architecture, focusing on the Intelligence Layer and highlighting the benefits and functionalities it provides to administrators and users of the edge-to-cloud continuum. It also describes in detail some experimental results to predict the CPU utilization of the nodes that build up the ICOS system. The purpose of this paper is to show the benefits of using ICOS with AI-subsystem and illustrate them through real experiments.
LOMOS: An AI-Based Runtime Security Monitoring System Fit for the Cloud Continuum
Given the challenges faced by various industries in the global digital transformation process, it is essential to perform detection of anomalies, consuming system logs collected and returning anomaly score, which should significantly enhance the visualization of vulnerabilities and improve the overall security posture of systems. This paper presents LOg MOnitoring System (LOMOS), a robust AI technology and methodology for anomaly detection on logs, tailored to adapt to new data sensitivity concerns. LOMOS facilitates the creation of informative metrics/variables with significant screening capabilities, addressing the critical need for real-time monitoring of stack conditions to fuel its self-healing mechanisms. The proposed system is designed to detect security related events and incidents within the deployed application environment and is deployable automatically, providing users with timely notifications about security episodes. In this paper, we demonstrate the advantages of this approach in the continuous detection of vulnerabilities, threats and malware in production infrastructures and during software development phases, appearing in the infrastructure when new services or features are added, or simply when new vulnerabilities are discovered in existing (outdated) services. By seamlessly integrating this novel transformer-based anomaly detection methodology with the cloud continuum, it facilitates a smooth and secure digital transformation process, ensuring a comprehensive adherence to evolving security requirements while supporting the dynamic nature of modern infrastructures.
Multi-component Application Mapping Across the Continuum
This paper presents a novel orchestrating mechanism for multi-component applications across the continuum. The orchestration mechanism provides an initial components-to-nodes mapping at launch time and, in case of anomaly detection at runtime, reschedules the application layout to meet the performance objectives within the application-specified constraints. The mapping engine, named Match Maker, is a core component inside the ICOS research project, an intelligent metaOS conceived to leverage the capabilities of the continuum. It is fed with abundant and accurate information collected at runtime from the telemetry component and enriched with additional information from the forecasting system generated by the intelligence layer. The novelty of the orchestrating mechanism is that the provided layout distributes the application components across the continuum, fulfilling the application requirements and constraints in the selected nodes, and considering the interdependence effects between selected nodes. In addition, the application can define policies for layout decisions based on performance and security, and policies for anomaly detection and remediation actions.
Placing Computational Tasks Within Edge-Cloud Continuum: A DRL Delay Minimization Scheme
In the rapidly evolving landscape of IoT-Edge-Cloud continuum (IECC), effective management of computational tasks offloaded from mobile devices to edge nodes is crucial. This paper presents a Distributed Reinforcement Learning Delay Minimization (DRL-DeMi) scheme for IECC task offloading. DRL-DeMi is a distributed framework engineered to tackle the challenges arising from the unpredictable load dynamics at edge nodes. It empowers each edge node to independently make offloading decisions, optimizing for non-divisible, latency-sensitive tasks without reliance on prior knowledge of other nodes’ task models and decisions. By framing the problem as a multi-agent computation offloading scenario, DRL-DeMi aims to minimize expected long-term latency and task drop ratio. Adhering to IECC requirements for seamless task flow within the Edge layer and between Edge-Cloud layers, DRL-DeMi considers three computation decision avenues: local computation, horizontal offloading to another edge node, or vertical offloading to the Cloud. Integration of advanced techniques such as long short-term memory (LSTM), double deep Q-network (DQN), and dueling DQN enhances long-term cost estimation, thereby refining decision making efficacy. Simulation results validate DRL-DeMi’s superiority over baseline offloading algorithms, showcasing reductions in both task drop ratio and average delay.
Navigating the Dynamic Heterogeneous Computing Sphere: The Role of EdgeHarbor as a Multi-edge Orchestrator
The term "Dynamic Heterogeneous Computing Sphere" is employed to delineate a computing paradigm that is heterogeneous, volatile, and highly dy- namic. This is the consequence of the incorporation of a substantial number of compute, storage, network, etc. resources, which are distributed across a spec- trum that encompasses both infrastructure (physical and virtual) and data. These resources are distributed across a range of locations, from the extreme IoT to the edge (far and near) and to the cloud. To deploy and execute a set of innovative vertical applications that are heterogeneous, it is necessary to consider the de- ployment of a suitable computing paradigm. The envisaged system requires the capacity for both resources and services to be elastic, that is, capable of being continuously shaped and moulded to support the specific needs of highly de- manding applications, both in terms of allocation and runtime. Moreover, soft- ware modules, which comprise vertical applications, should be partitioned intel- ligently into virtual elements (i.e., containers) to optimise their placement and subsequent execution. This can be achieved by considering aspects such as per- formance and eco-efficiency aspects. The necessity for ad hoc resource and ser- vice shaping is growing in accordance with the prevailing tendencies towards the softwareisation of systems management. This has been further driven by the con- cept of disaggregation and the advent of new ultra-real-time services with high performance demands (X-AR, holo, metaverse, etc.). This paper introduces the EdgeHarbor orchestrator service, an open-source, multi-edge management sys- tem for a dynamic heterogeneous computing sphere. The work has been sup- ported by the European Union's Horizon research and innovation programme un- der grant agreement ICOS (www.icos-project.eu), grant number 101070177.
Towards a Functional Continuum Operating System – ICOS MetaOS
This article presents the intermediate results of the ICOS project, which aims to create a meta-operating system for cloud continuum, and describes the adaption of ICOS in two (out of four) project’s pilot Use Cases. For each described scenario, the paper provides insight into the specific key architectural features, identifies the expected benefits, and provides some initial validation results. The purpose of this research is to highlight the advantages of using a metaOS in real scenarios as well as to illustrate the ease of integration into ICOS.
Dual-Link Data Resilient Edge-to-cloud Communication Framework for Agricultural Robots
Reliable and high-throughput communication between field robots and cloud services remain a key challenge in precision agriculture, where remote rural areas often lack consistent high-bandwidth connectivity. In this work, we introduce a new dual-link edge-to-cloud data transfer framework that combines long-range Low-Power Wide Area Networking (LPWAN) for essential control and monitoring with IEEE 802.11 Wi-Fi that carries bulk data over a Zenoh protocol. In addition, a data router dynamically switches the robot between ‘Transfer Mode’, in which sensor streams and imagery data are being forwarded via Wi-Fi, and ‘Storage Mode’, in which data are locally recorded in Robotic Operating System (ROS) 2 bags to prevent loss when connectivity degrades. To preemptively detect Wi-Fi link failures and issue routing instructions to the data router, an onboard anomaly detection node monitors heartbeat timing using a machine learning-based algorithm, namely the XGBoost model. Field trials demonstrate that (1) Wi-Fi transfers maintain sub-100 ms latency within 240 m of the gateway, (2) Long Range (LoRa) communication persists reliably beyond 350 m with ≈0.1 s latency, (3) the router achieves an average of 0.8 s overlap when entering Storage Mode, and (4) the anomaly detector successfully flags link degradation ahead of an outage. Our framework scales to multi-robot deployments via ROS 2 namespaces and Zenoh multicast, laying the groundwork for resilient swarm operations in rural environments.
This project has received funding from the European Union’s HORIZON research and innovation programme under grant agreement No 101070177.