The rapid expansion of the cloud-native landscape, particularly the growing number of orchestration distributions for cloud and edge computing, presents a critical challenge for systematic tool selection. Existing empirical studies, including prior comparative analyses, are inherently constrained by specific benchmarks, hardware, or limited tool subsets, yielding fragmented knowledge that fails to address the holistic requirement of "Which system is most suitable for my use case?". To address this fragmentation, we introduce Di-Select, a criteria-driven conceptual framework for evidence-based orchestration selection. The framework was derived through a two-tier methodology: a Structured Literature Review followed by Grounded Theory coding (Open, Axial, and Selective). This synthesis produced seven high-order evaluation constructs (Performance, Security, Connectivity, Reliability, Ecosystem, Maintainability, and Resource and Cost) and fifteen causal propositions (P1–P15) that formalise the trade-offs between criteria, such as the inverse relationship between default security hardening and resource consumption or recovery time. The coding was verified through independent double-coding with inter-coder agreement of Cohen’s κ = 0.84 . Di-Select operates through a structured workflow that integrates a Knowledge Base (Concept Map + Data), a granular User Questionnaire, and a Decision and Scoring Engine. By translating user priorities into quantifiable weights, the framework performs systematic trade-off and scenario analysis, enabling transparent and rationale-driven selection. The current empirical grounding comes mainly from Kubernetes and its lightweight variants, although the constructs and scoring methodology are not tied to any single technology and can be extended to other orchestrators. Demonstrations show how Di-Select shifts recommendations based on context, favoring specialized edge solutions (e.g., KubeEdge) when connectivity and cost are critical, and robust cloud platforms (e.g., Full Kubernetes) when performance and ecosystem support dominate. Internal explanatory adequacy is confirmed through logical path-tracing of the Concept Map. External validation across heterogeneous, real-world deployments remains future work and is the focus of the validation roadmap outlined in the Discussion.

The rapid increase in the number of connected devices has created a vast computational infrastructure. While intelligent applications (e.g., VR/AR, autonomous driving, and AI assistants) demand substantial processing power, they are typically executed only in short bursts. As a result, devices remain idle for significant periods, presenting an opportunity to utilize their unused computational resources. Computation offloading has emerged as a viable solution, where resource-constrained devices leverage edge or cloud infrastructure for intensive tasks. More recently, horizontal offloading has gained attention as a complementary approach, enabling devices within the same layer to collaborate and utilize idle resources more efficiently. In this survey, we examine current research on computation offloading in collaborative computing systems that span from horizontal offloading among end devices to more complex systems involving horizontal as well as vertical offloading to edge/cloud systems. Our examination focuses on architectures, collaborative characteristics, and offloading dynamics, such as task dependency, mobility, and multi-hop offloading. We begin by introducing edge computing paradigms and key concepts in computation offloading and collaborative computing. The offloading approaches are then classified into classical, heuristic, metaheuristic, and machine learning methods, further divided into centralized, decentralized, and distributed categories. These approaches are analyzed and compared based on collaborative system, architecture, offloading dynamics, and evaluation methodology. Finally, we have identified open challenges and emphasized the need to unify research across edge computing paradigms, alongside a focus on more complex collaborative systems that resemble real-world systems, and the development of standardized evaluation frameworks as key directions for future work.

We present a engineering case study on the design and evolution of a cloud-native MLOps pipeline in an industrial setting using Databricks. Motivated by the operational and lifecycle challenges of deploying machine-learning models at scale, the project adopted a platform-centric approach to automation, reproducibility, and scalable operation. The pipeline integrates versioned data management, experiment tracking and registry-based promotion, CI/CD for ML, and environment isolation to bridge experimentation and production. Security and data-privacy constraints are operationalized through workspace isolation, role-based access control, and secrets management integrated into deployment automation. Our methodological approach entailed an engineering case-study design, incorporating triangulation across development artifacts, CI/CD records, and collaborative design episodes. The paper delivers three contributions: (i) a platform-grounded reference architecture documenting the design decisions, trade-offs, and deliberate deviations from vendor guidance that shaped the final implementation; (ii) collaboration practices that align roles across DataOps, ModelOps, and DevOps showing how shared artifacts and platform constraints structure cross-role coordination; and (iii) recurring implementation patterns and practitioners lessons that are capability-oriented and transferable beyond the specific platform. The results provide practitioners with empirical evidence on how MLOps automation is realized and constrained in practice, filling a gap that vendor documentation and conceptual frameworks alone do not address.

A challenge in designing IoT and Cyber-Physical Systems (CPS) is the fact that their structure is often rigid. Even in the cloud-edge continuum, it is common that some functions are statically located at a certain node, and their relocation would require additional implementation effort. Furthermore, computing capacity at the network edge may be low, which poses challenges to using machine learning (ML) models. In this paper, we study the performance limitations associated with edge nodes, if allocation of computations was a dynamic process instead of predefined one. Thus, we follow the paradigm of liquid software. We first explain how ML models are deployed and used in such dynamic context. Then, we benchmark different hardware and software implementations that are relevant in the CPS context. Finally, we analyze what each result means in practice.

Complex logistics systems may require the cooperation of a variety of subsystems, such as robots, autonomous vehicles, and human-operated devices. These so-called mixed-fleet systems are found in logistics use cases, which demand high flexibility. Hence, the choreography that orchestrates the involved software systems, spanning edge to cloud, must be adaptable and reconfigurable. Enabling to add or remove subsystems flexibly during runtime requires a strong decoupling, which is found in multi-agent systems. In this paper, we explore a model-driven engineering process for service choreographies of flexible, heterogeneous, and autonomous mixed-fleet systems. Each complex process is decomposed into services and tasks, which can then be flexibly assigned to available actors. The resulting layered service-oriented architecture is realized as an event-based system. We define requirements for the modeling of processes, services, interactions, and event objects, before evaluating different modeling languages based on their applicability for each layer, i.e., BPMN, IEC 61499, and SysML. We then demonstrate and evaluate our architecture using a port logistics use case. The results show that the three languages are suitable candidates for modeling event-based process models, although they have varying advantages and disadvantages. Overall, their diagrams can be used to capture service choreography models for decentralized systems. Future work will investigate how these models can be validated comprehensively and used for system implementation.

Federated Learning (FL) is a collaborative learning paradigm in which multiple clients train a shared global model without exchanging data. Clients communicate only model updates with a central aggregator. In parallel, Machine Learning Operations (MLOps) streamline the development, deployment, and monitoring of ML systems, while their extension, Federated Learning Operations (FLOps), aims to bring the operational discipline to decentralized and privacy-sensitive settings.This study presents a systematic mapping study (SMS) on FL and FLOps, and clarifies foundational concepts and uncovers new perspectives within this evolving field. We focus on FLOps and FL features: architecture, communication and aggregation models. First research question (RQ) focuses on prevalent FL computing architectures. Second RQ covers data transfer between FL components. Third RQ determines breadth of FLOps application in the scientific literature. Fourth RQ identifies distinct approaches to global model aggregation.Our analysis reveals that Edge-based local training with Cloud-based aggregation is the most adopted architecture, combining Edge privacy and responsiveness with Cloud computational capacity. Communication is enabled through lightweight protocols such as Message Queuing Telemetry Transport (MQTT), but protocol choice depends on constraints. Notably, FLOps remains a rarely addressed topic, indicating a substantial gap in end-to-end support for FL pipelines. Federated Averaging (FedAvg) is the most employed aggregation approach, valued for its simplicity and effectiveness with heterogeneous data. These findings expose critical research gaps in architectural diversity, protocol selection, lifecycle integration and adaptive aggregation, highlighting the need for more cohesive and scalable FL system design in future work.

Localization within a known environment is a crucial capability for mobile robots. Simultaneous Localization and Mapping (SLAM) is a prominent solution to this problem. SLAM is a framework that consists of a diverse set of computational tasks ranging from real-time tracking to computation-intensive map optimization. This combination can present a challenge for resource-limited mobile robots. Previously, edge-assisted SLAM methods have demonstrated promising real-time execution capabilities by offloading heavy computations while performing real-time tracking onboard. However, the common approach of utilizing a client-server architecture for offloading is sensitive to server and network failures. In this article, we propose a novel edge-assisted SLAM framework capable of self-organizing fully distributed SLAM execution across a network of devices or functioning on a single device without connectivity. The architecture consists of three layers and is designed to be device-agnostic, resilient to network failures, and minimally invasive to the core SLAM system. We have implemented and demonstrated the framework for monocular ORB SLAM3 and evaluated it in both fully distributed and standalone SLAM configurations against the ORB SLAM3. The experiment results demonstrate that the proposed design matches the accuracy and resource utilization of the monolithic approach while enabling collaborative execution.

Machine learning operations (MLOps) has rapidly evolved from a marginal concept to a pivotal consideration for any enterprise implementing ML at scale. Cloud providers have rushed to fill this gap, each offering its own take on what MLOps should be. This article compares the MLOps offerings of Amazon Web Services (AWS), Azure, Google Cloud Platform (GCP), and Databricks. Their analysis reveals the diverse perspectives guiding each vendor’s approach, assisting practitioners in identifying suitable tradeoffs for their specific contexts. — Davide Taibi

Edge computing environments increasingly rely on lightweight container orchestration platforms to manage resource-constrained devices. This paper provides an empirical analysis of five lightweight kubernetes distributions (KD)(k0s, k3s, KubeEdge, OpenYurt, and Kubernetes (k8s)) focusing on their performance and resource efficiency in edge computing scenarios. We evaluated key metrics such as CPU, memory, disk usage, throughput, and latency under varying workloads, utilizing a testbed of Intel NUCs and Raspberry Pi devices. Our results demonstrate significant differences in performance: k3s exhibited the lowest resource consumption, while k0s and k8s excelled in data plane throughput and latency. Under heavy stress scenarios, k3s and k0s accomplished the same workloads faster than the other distributions. OpenYurt offered balanced performance, suitable for hybrid cloud-edge use cases, but was less efficient in terms of resource usage and scalability compared to k0s, k3s and k8s. KubeEdge, although feature-rich for edge environments, exhibited higher resource consumption and lower scalability. These findings offer valuable insights for developers and operators selecting appropriate KD based on specific performance and resource efficiency requirements for edge computing environments.

Background: The rapid growth of IoT and edge devices demands low-latency, scalable computing solutions. Traditional cloud architectures introduce bottlenecks, while serverless computing, particularly Function-as-a-Service (FaaS), offers an event-driven, on-demand execution model. However, existing FaaS platforms rely on centralized orchestration, leading to cold start delays, resource inefficiencies, and limited adaptability in edge environments. This study examines containerized workload management for serverless edge computing, evaluating orchestration, choreography, and offloading techniques to improve function execution efficiency and system scalability. Objective: The aim of this work is to conduct a systematic review the state-of-the-art practices in orchestration and offloading techniques for containerized environments. Method: We conducted a Multivocal Literature Review, selecting 94 works from a pool of 771. Results: The offloading strategies rely on a central orchestrator and on the use of well-established tools such as Kubernetes and Docker. Few studies propose alternatives, highlighting a lack of diversity in offloading tools. These studies underscore a significant lack of diversity in the tools used for offloading in containerized systems. Conclusion: We have identified key strengths and limitations of the existing approaches. These limitations are primarily related to response time, performance, and resource utilization.

The use of unmanned aircraft vehicles (UAVs) for various tasks and services has increased significantly in recent years. One of the most important application areas is first-response missions, which UAVs can enhance by speeding up both resource transportation and data acquisition. To support these critical operations, this paper introduces a UAV digital platform architecture. The digital platform integrates autonomous air operations, U-space management, and aerial services to meet stakeholders’ requirements and support the development of a multitenant UAV infrastructure. The architecture of the platform is validated using real-world urgent scenarios, including the delivery of a defibrillator and a search operation. These scenarios prove the microservices-based platform effectively performs the required functions, such as mission optimization, resource allocation, fleet management, contextual information sharing, and security assurance. The architecture can also be expanded to other drone services than the first response.

Modern logistics systems aim to leverage digital technologies and may integrate autonomous components to increase efficiency and flexibility. In so-called mixed-fleet systems, human workers, manually operated machines, and autonomous machines collaboratively work towards a common goal. The subsystems are loosely coupled and can be reconfigured flexibly, leading to a change in behavior. Modeling this behavior requires flexible designs for model-driven systems engineering. Serviceoriented architectures can help focus on defining the expected behavior, independently of the involved actors. Additionally, eventbased communication mechanisms can decouple interactions between subsystems. This paper explores the use of SysML v2 for modeling a service-oriented architecture of mixed-fleet systems. Based on an available set of requirements, suitable SysML v2 modeling elements are identified that can describe services, events, and service choreographies. We use the described concepts to create a SysML v2 model of a mixed-fleet harbor logistics use case. Based on this model, we demonstrate how business processes can be composed of reusable services and how requirements and verification can be integrated to ensure correctness of behavior. The results show that SysML v2 meets key requirements for service-oriented architectures and enables separating service definitions, actors, and verification elements. Reusable modeling patterns were applied to support scalability, and enable traceability within the model across actors and services. Furthermore, domain-specific constraints and requirements were composed into modeling elements using formal mechanisms to ensure that they are not only documented, but actively connected to the model.

Background: The growing demand for efficient and deployable language models has led to increased interest in Small Language Models (SLMs). However, existing research remains fragmented, lacking a unified lifecycle perspective. Objective: This study aims to define a comprehensive lifecycle framework for SLMs by synthesizing insights from academic literature and practitioner sources. Method: We conducted a comprehensive survey of 36 works, analyzing and categorizing lifecycle-relevant techniques. Results: We propose a modular lifecycle model structured into main, optional, and cross-cutting components. The model captures key interconnections across stages, supporting method reuse, co-adaptation, and lifecycle-awareness. Conclusion: Our framework provides a coherent foundation for developing and maintaining SLMs, bridging theory and practice, and guiding future research and tool development.

The increasing demand for real-time data processing in Internet of Things (IoT) devices has elevated the importance of edge computing, necessitating efficient and secure deployment of applications on resource-constrained devices. Kubernetes and its lightweight distributions (k0s, k3s, KubeEdge, and OpenYurt) extend container orchestration to edge environments, but their security, reliability, and maintainability have not been comprehensively analyzed. This study compares Kubernetes and these lightweight distributions by evaluating security compliance using kube-bench, simulating network outages to assess resiliency, and documenting maintainability. Results indicate that while k3s and k0s offer superior ease of development due to their simplicity, they have lower security compliance compared to Kubernetes, KubeEdge, and OpenYurt. Kubernetes provides a balanced approach but may be resource-intensive for edge deployments. KubeEdge and OpenYurt enhance security features and reliability under network outages but increase complexity and resource consumption. The findings highlight trade-offs between performance, security, resiliency, and maintainability, offering insights for practitioners deploying Kubernetes in edge environments.

In distributed edge computing, effective resource management is key. While orchestration offers centralized control, choreography provides a decentralized alternative more suited to dynamic, resource-constrained environments-especially within emerging serverless paradigms. This study investigates choreography techniques for serverless edge computing, analyzing coordination strategies, communication models, and application domain. We conducted a Multivocal Literature Review (MLR), selecting 12 peer-reviewed studies from 332 sources (200 gray, 122 white literature). Three main choreography approaches emerged: structured, scheduling-based, and hybrid. End-to-end communication was twice as common as broadcasting. IoT was the most prevalent domain, consistent with the dominant use of lightweight edge devices. Choreography offers performance and flexibility benefits but remains underexplored. Key limitations include resource management challenges, underscoring the need for further research into scalable, decentralized coordination models for the edge.

Complex processes require the cooperation of a variety of subsystems, such as robots, autonomous vehicles, and human-operated devices. These so-called mixed-fleet systems are found in logistics and production use cases, which also demand a high flexibility. Hence, the choreography that orchestrates the involved systems must be adaptable and reconfigurable. Enabling to add or remove subsystems flexibly during runtime requires a strong decoupling, which is found in multi-agent systems. In this paper, we explore a model-driven engineering process for service choreographies of flexible, heterogeneous, and autonomous mixed-fleet systems. Each complex process is decomposed into services and tasks, which are flexibly assigned to resources. The resulting layered service-oriented architecture is realized as an event-based system. We define requirements for modeling services, tasks, and events and evaluate different modeling language based on their applicability for each layer, i.e., BPMN, SysML/UML, and IEC 61499. We demonstrate and evaluate our architecture using a logistics use case scenario. The results show that these languages are suitable candidates for modeling event-based process models and that the diagrams can be used to capture service choreography models for decentralized systems. Future work will investigate how these models can be validated comprehensively and used for system implementation.

Edge-to-cloud computing is an emerging paradigm for distributing computational tasks between edge devices and cloud resources. Different approaches for orchestration, offloading, and many more purposes have been introduced in research. However, it is still not clear what has been implemented in the industry. This work aims to merge this gap by mapping the existing knowledge on edge-to-cloud tools by providing an overview of the current state of research in this area and identifying research gaps and challenges. For this purpose, we conducted a Multivocal Literature Review (MLR) by analyzing 40 tools from 1073 primary studies (220 PS from the white literature and 853 PS from the gray literature). We categorized the tools based on their characteristics and targeted environments. Overall, this systematic mapping study provides a comprehensive overview of edge-to-cloud tools and highlights several opportunities for researchers and practitioners for future research in this area.