[Paper Review] The SPEC-RG Reference Architecture for the Compute Continuum

Matthijs Jansen, Auday Al-Dulaimy+3 more|arXiv (Cornell University)|Jul 9, 2022

IoT and Edge/Fog Computing4 citations

TL;DR

This paper proposes the SPEC-RG Reference Architecture for the compute continuum, unifying fragmented edge, fog, and cloud computing models into a cohesive framework. It enables systematic workload reasoning through a deployment and benchmarking framework with a first-order analytical model, validated on deep learning and industrial IoT workloads, offering open-source tools for end-to-end performance and resource management analysis across the continuum.

ABSTRACT

As the next generation of diverse workloads like autonomous driving and augmented/virtual reality evolves, computation is shifting from cloud-based services to the edge, leading to the emergence of a cloud-edge compute continuum. This continuum promises a wide spectrum of deployment opportunities for workloads that can leverage the strengths of cloud (scalable infrastructure, high reliability) and edge (energy efficient, low latencies). Despite its promises, the continuum has only been studied in silos of various computing models, thus lacking strong end-to-end theoretical and engineering foundations for computing and resource management across the continuum. Consequently, developers resort to ad hoc approaches to reason about performance and resource utilization of workloads in the continuum. In this work, we conduct a first-of-its-kind systematic study of various computing models, identify salient properties, and make a case to unify them under a compute continuum reference architecture. This architecture provides an end-to-end analysis framework for developers to reason about resource management, workload distribution, and performance analysis. We demonstrate the utility of the reference architecture by analyzing two popular continuum workloads, deep learning and industrial IoT. We have developed an accompanying deployment and benchmarking framework and first-order analytical model for quantitative reasoning of continuum workloads. The framework is open-sourced and available at https://github.com/atlarge-research/continuum.

Motivation & Objective

• To address the lack of end-to-end theoretical and engineering foundations for resource management and workload distribution across the cloud-edge-endpoint continuum.
• To unify isolated computing models—such as fog, mist, edge, and mobile cloud computing—into a single, coherent reference architecture.
• To provide developers with a systematic framework for reasoning about workload offloading, resource allocation, and performance across heterogeneous continuum deployments.
• To develop a deployable, open-source framework that supports emulation of diverse continuum resources and enables quantitative benchmarking.
• To establish a first-order analytical model for performance reasoning that complements the deployment framework and supports rapid configuration iteration.

Proposed method

• Systematically analyzed 17 existing compute continuum models to identify overlapping concerns and design space characteristics.
• Selected five representative models—mist, edge, multi-access edge, fog, and mobile cloud computing—to cover distinct parts of the continuum design space.
• Designed a unified reference architecture that abstracts commonalities across models, enabling end-to-end reasoning on workload distribution and resource management.
• Implemented a deployment and benchmarking framework supporting virtual machines, containers, and multiple resource managers (e.g., Kubernetes, KubeEdge) across cloud, edge, and endpoint devices.
• Developed a first-order analytical model for performance reasoning, integrated with the benchmarking framework to enable rapid configuration evaluation.
• Enabled customization of compute, network, and operating services, supporting full-stack emulation of continuum deployments.

Experimental results

Research questions

• RQ1 How can disparate computing models like fog, edge, mist, and mobile cloud computing be unified under a single conceptual framework for the compute continuum?
• RQ2 What are the core commonalities and distinguishing characteristics across existing continuum models that enable a systematic design space exploration?
• RQ3 How can a reference architecture support end-to-end reasoning about workload offloading, resource management, and performance across the cloud-edge-endpoint continuum?
• RQ4 To what extent can a unified deployment and benchmarking framework improve the evaluation and optimization of continuum workloads?
• RQ5 Can a first-order analytical model, coupled with a configurable framework, enable efficient and quantitative reasoning for continuum application deployment?

Key findings

• The SPEC-RG Reference Architecture successfully unifies 17 diverse computing models under a single, coherent framework, revealing significant overlap in concerns despite isolated presentations.
• The proposed deployment and benchmarking framework supports full-stack emulation of cloud, edge, and endpoint resources, including configurable resource managers and operating services, enabling comprehensive experimentation.
• The framework outperforms existing tools like Fogify and MockFog by supporting all continuum computing models and enabling fine-grained control over compute and network resources.
• The first-order analytical model enables rapid performance reasoning across diverse configurations, reducing the need for exhaustive trial-and-error deployment tuning.
• Validation on deep learning and industrial IoT workloads demonstrates the framework’s utility in analyzing real-world continuum workloads with measurable performance and resource utilization insights.
• The open-source framework and model are actively developed and available at https://github.com/atlarge-research/continuum, supporting ongoing research and tool integration.