Towards a Definition of Computing Continuum

Hacia una definición del Computing Continuum

Pablo Josue Rojas Yepes sthongurdt@hotmail.com

Universidad Industrial de Santander, Colombia

 https://ror.org/04n6qsf08

Carlos Jaime Barrios Hernández

Universidad Industrial de Santander, Colombia

 https://ror.org/04n6qsf08

Oscar Carrillo

INSA Lyon, Francia

 https://ror.org/050jn9y42

Frédéric Le Mouël

INSA Lyon, Francia

 https://ror.org/050jn9y42

Esta obra está bajo una Licencia Creative Commons Atribución-NoComercial 4.0 Internacional.

Since the early 21st century, computational capabilities have addressed significant human needs, leading to the development of methods and tools to ensure continuity across various domains such as data, infrastructure, operations, and development. A Computing Continuum system is designed to manage continuous data streams and enable real-time processing for ultra-scale computing applications. Different definitions of Computing Continuum exist, such as the one in [1], which focuses on integrating resources from the edge to the cloud to support dynamic, data-driven applications, or the approach in [2], which introduces a “Markov Blanket”-based methodology to manage distributed systems, overcoming current limitations and establishing a more flexible framework.

The contemporary computing landscape is dynamic, heterogeneous, and vast, necessitating a paradigm shift for effective management and orchestration of applications across this complex ecosystem. The Computing Continuum extends beyond the traditional cloud model, incorporating entities like edge devices, fog nodes, and the IoT, all contributing to the continuum by providing analysis, processing, storage, and data generation capabilities.

Architecturally, a Computing Continuum system encompasses large-scale, complex, and distributed systems, combining parallel computing, heterogeneity, scalability, and large data volumes. Applications within this continuum often scale by orders of magnitude, similar to ultra-scale computing systems [3]. However, while ultra-scale computing focuses on extending individual systems, the Computing Continuum seeks to unify computing resources into a seamless continuum. The shared architectural features suggest that ultrascale computing systems can be components of a Computing Continuum.

The primary objective in application deployment within the continuum is to leverage its extensive capabilities. This includes the low latency and proximity benefits of edge computing and the scalability of cloud data centers, as well as the integration of resources across edge, core, and data paths to support dynamic, flexible, data-driven workflows. Effective orchestration is crucial in this context.

Orchestration systems automate the seamless delivery of applications across diverse cloud environments, ensuring that Quality of Service (QoS) objectives are met. These systems manage tasks like resource selection, deployment, monitoring, and runtime control. The rise of DevOps has significantly transformed software development and delivery, emphasizing collaboration and agility. The synergy between orchestration and DevOps strengthens the Computing Continuum by fostering efficiency, flexibility, and resilience in managing applications across diverse architectures.

This work is divided into four sections. The first one will address topics related to the concept of the Computing Continuum and related works (the bibliography is extensive, so individual articles may be omitted). The second will present the contributions of orchestration and DevOps CI/CD to the Computing Continuum. The third will discuss how the contributions of Synergy into Apps, Orchestrations, Devops and infrastructure section relate to those outlined in the concept of Computing Continuum and related works section. Finally, the fourth section will present the conclusions of this work. It should be noted that this study is an exploration of the extensive theory surrounding the concept of the Computing Continuum, the characterization of the challenges to be faced, and proposals for potential solutions based on well-defined topics (Orchestration and DevOps). This is why we do not include use cases in this work.

The Computing Continuum has evolved over nearly a decade, propelled by advancements such as diverse cloud services, the surge in computing [4] and data manipulation [5] capabilities in fog and edge devices, the democratization of IoT through open hardware initiatives, and the adoption of DevOps culture. DevOps integrates repositories, development, deployment, and operations with orchestrators, facilitating the efficient management of large-scale data across multiple architectures. Conceptually, the Computing Continuum unifies various computing resources with distinct capabilities and characteristics.

Fog and cloud computing address challenges like quality of service [6] and security [7] in SDN and NFV environments, enabling applications in sectors such as agriculture 4.0 [8]. The digital transformation and IoT require new strategies, highlighting the importance of time-aware data spaces for scalability [9]. Fog computing extends cloud services to the network edge, supporting IoT, 5G, and AI applications [10]. The cloudto-things continuum, integrated with technologies like iCloudFog [11, 12], manages latency-sensitive IoT data and addresses resource management and security challenges, facilitating real-time applications like tsunami warning systems [13].

Figure 1
Representation of a Computing Continuum Concept. Own elaboration

Figure 1. Representation of a Computing Continuum Concept. Own elaboration.

Optimizing resources in IoT-fog-cloud environments is crucial [14]. Rule-based resource matching [15] and emulation approaches [16, 17] enhance service placement, maximizing performance in computing continua. Serverless computing [18] and dynamic function distribution offer scalable solutions for complex applications [19]. Tools like Continuum Deployer [20] and models like CODA [21] optimize application and service allocation, considering performance, carbon footprint [22], and quality of service. Innovative approaches like EPOS Fog [23] and the Markov Blanket concept [24] improve efficiency in distributed environments [25].

Key features of the Computing Continuum include minimizing latency [21] and optimizing data processing [15] by distributing computational tasks across edge, fog, and cloud levels [14], enhancing overall system performance [25, 26, 27].

A conceptual architecture of a Computing Continuum system, as depicted in Fig. 1, identifies different elements and levels, illustrating their relationships across varying scales, while considering application behavior and data characteristics. This representation extends beyond the conventional cloud model, incorporating edge, fog, and IoT entities that contribute to the continuum by offering analysis, processing, storage, and data generation capabilities [6].

This evolution, underscored by integrating DevOps with these entities, highlights the dynamic nature of modern computing paradigms [11]. Central to this concept is the ACDC (Application Choose Device Carefully) proposal [28], which emphasizes the flexibility and scalability of deploying general-purpose systems across diverse devices. Orchestrators play a crucial role in automating application delivery across environments, though orchestrating this continuum requires solutions that ensure efficient execution and robust security, which will be addressed in the next section.

The integration of orchestration systems and DevOps practices within the Computing Continuum as shown in Fig. 2 represents a powerful synergy that addresses many of the challenges associated with managing heterogeneous and distributed computing environments. This synergy is highlighted in Table 1, which outlines the relationship between various computing architectures and their performance, offering a clear view of how these technologies interact to create a more efficient and resilient computing infrastructure.

TABLE 1. Relationship between Computing Continuum References (Architecture and Performance)

TABLE 1
Relationship between Computing Continuum References (Architecture and Performance)

Orchestration systems such as Kubernetes and Docker Swarm play a pivotal role in managing the complex landscape of the Computing Continuum by automating the deployment, scaling, and management of applications across diverse environments, from the cloud to the edge. These systems are essential for ensuring that (QoS) goals are met, particularly in dynamic and distributed computing environments where resource heterogeneity and geographical distribution present significant challenges [29]. The ability of orchestration systems to handle resource selection, deployment, and runtime control is crucial for optimizing resource utilization and minimizing latency across the continuum [30, 34]. This is particularly important in scenarios where real-time processing and fault tolerance are critical, as these systems can dynamically adjust to changing conditions and maintain service continuity.

The integration of DevOps culture within the orchestration framework further enhances the effectiveness of these systems. DevOps practices emphasize automation, collaboration, and continuous improvement, breaking down traditional silos between development and operations teams and fostering a culture of shared responsibility. This alignment between orchestration and DevOps enables organizations to achieve greater efficiency, flexibility, and resilience in their computing infrastructure [64]. By automating the integration, testing, and delivery of software artifacts, DevOps ensures that development and operations teams work seamlessly together to deliver and maintain applications across the Computing Continuum [44, 66, 67].

The convergence of orchestration and DevOps within the Computing Continuum also enhances security by automating security measures and enforcing policy-based deployment and configuration. This is critical in environments where applications and data span multiple architectures, including cloud, edge, and IoT devices. The security mechanisms provided by orchestration systems help mitigate the challenges of volatility, mobility management, and security, ensuring that the Computing Continuum remains secure and resilient against various threats [44, 69].

Despite the clear benefits, implementing an integrated Computing Continuum with DevOps practices is not without challenges. The complexity of coordinating across multiple administrative domains, managing resource volatility, and ensuring efficient and secure monitoring are significant obstacles. Additionally, the shortage of skilled DevOps professionals and organizational resistance to change can hinder the adoption and success of these practices. Overcoming these challenges requires effective change management strategies and robust training programs to equip teams with the necessary skills [42, 6, 69, 72].

The synergy between orchestration systems and DevOps practices is essential for navigating the complexities of the Computing Continuum. As illustrated in Table 1, this synergy enhances the efficiency, scalability, and security of computing environments, enabling organizations to adapt quickly to evolving demands and technological advancements. Future research should focus on developing more adaptive and secure orchestration solutions, as well as exploring strategies for training and organizational change that facilitate the successful integration of DevOps within the Computing Continuum.

In conclusion, the Computing Continuum represents a transformative and integrated ecosystem that encompasses the entire spectrum of computing resources, from the network edge to the cloud. This paradigm integrates a variety of computing architectures—cloud, fog, edge, and IoT—into a unified system designed to deliver scalable, efficient, and resilient solutions. The evolution of this concept is driven by advances in cloud services, the increasing adoption of edge and fog computing, the rise of open hardware initiatives, and the incorporation of DevOps practices.

At its core, the Computing Continuum aims to reduce latency and optimize data processing by distributing computational tasks across multiple layers. This strategic approach enhances resource utilization and system performance by enabling the simultaneous execution of workloads on edge, fog, and cloud tiers. The seamless integration of these architectures, supported by containerization technologies and orchestration, provides the flexibility and agility necessary to meet diverse application requirements.

Furthermore, the integration of DevOps within the Computing Continuum strengthens its capabilities by fostering collaboration and automation. DevOps practices dissolve traditional barriers between development and operations teams, enabling more efficient and scalable application deployment across heterogeneous environments. This synergy addresses critical challenges such as federated coordination, volatility management, and security, thereby ensuring robust and adaptable computing infrastructures.

The Computing Continuum signifies a pivotal shift in computing paradigms, expanding beyond traditional cloud models to incorporate edge, fog, and IoT devices. By clearly defining the elements, contexts, and relationships within this ecosystem, the Computing Continuum emerges as a scalable and distributed computing system, capable of supporting a wide range of applications with varying demands. However, the orchestration of such a complex and heterogeneous system presents challenges, including managing resource capabilities, data diversity, resilience, mobility, and the scalability of applications. Overcoming these challenges is crucial to fully unlocking the potential of the Computing Continuum in the modern computing landscape.

In essence, the Computing Continuum is an abstract that offers a unifying approach to computing resources. By integrating diverse architectures and leveraging innovative technologies, it can deliver scalable, efficient, and resilient solutions. What truly distinguishes the Computing Continuum is its focus on minimizing latency and optimizing data processing through the simultaneous execution of tasks across multiple computing layers, a capability made possible by containerization technologies and DevOps practices.

It is with great pleasure that we acknowledge the invaluable contribution of Collaboration France-Colombie en Technologies Avancées de l’Information et la Communication (CATAÏ) in facilitating the collaboration between the Universidad Industrial de Santander (UIS) and Inria. Your support has not only enabled us to achieve our research goals but has also strengthened the ties between our two institutions.

Pablo Josue Rojas Yepes contributed to the roles of Investigation, Methodology, and Writing – Original Draft. He led the methodological design, executed the experiments, and prepared the initial manuscript.

Carlos Jaime Barrios Hernández participated in Validation, Project Administration, Supervision, and Funding Acquisition. He ensured the quality of the results, managed project activities, and secured the necessary financial resources.

Oscar Carrillo performed roles in Validation, Formal Analysis, Supervision, and Writing – Review & Editing. His contributions included detailed data analysis, academic supervision of the work, and critical review of the manuscript.

Frédéric Le Mouël participated in Validation, Conceptualization, Project Administration, and Supervision. His contributions involved defining the conceptual approach, securing funding, managing the project, and providing strategic guidance.

Authors declare they have no conflicts of interest in this research.

redalyc-journal-id: 7261