GMB-ECC: Guided Measuring and Benchmarking of the Edge Cloud Continuum

TL;DR

The paper addresses the challenge of measuring and benchmarking energy efficiency across the heterogeneous edge-cloud continuum. It presents GMB-ECC, a measurement and benchmarking framework that incorporates a precision parameter to adapt to diverse environments, implemented via a three-step methodology (State Representation, Energy Efficiency Analysis, Benchmarking). Key contributions include a weighted DAG state representation, cross-state graph merging, an efficiency gap relative to a theoretical optimum, Rosenblatt-inspired model fitness, and standardized benchmarking categories, demonstrated in an autonomous intra-logistics use case with a reported 12% reduction in total vehicle energy and substantial DVFS- and transmission-related savings. The work delivers a scalable, adaptable tool for practitioners to identify optimization opportunities and drive energy-efficient operations across edge, fog, and cloud layers. The practical impact lies in enabling sustainable, cost-effective deployment of heterogeneous edge-cloud systems across real-world industrial scenarios.

Abstract

In the evolving landscape of cloud computing, optimizing energy efficiency across the edge-cloud continuum is crucial for sustainability and cost-effectiveness. We introduce GMB-ECC, a framework for measuring and benchmarking energy consumption across the software and hardware layers of the edge-cloud continuum. GMB-ECC enables energy assessments in diverse environments and introduces a precision parameter to adjust measurement complexity, accommodating system heterogeneity. We demonstrate GMB-ECC's applicability in an autonomous intra-logistic use case, highlighting its adaptability and capability in optimizing energy efficiency without compromising performance. Thus, this framework not only assists in accurate energy assessments but also guides strategic optimizations, cultivating sustainable and cost-effective operations.

Figures (9)

• Figure 1: Energy consumption data flow. The energy is measured on the autonomous vehicles and sent to edge computing via the network. The edge computing and network energy is derived from utilization and measurement. All data is sent to the cloud for further processing and analysis.
• Figure 2: Software-Hardware Stack. Mapping of software processes to underlying hardware resources in the edge-cloud continuum.
• Figure 3: Energy efficiency graph. Showing components and their interrelations within the system, resulting in the composite energy efficiency curves for the whole edge-cloud continuum example sketched in Figure \ref{['fig:service-energy-measurement-model']}.
• Figure 4: Measurement and validation process. Measurements at different layers are compared and validated to ensure accuracy.
• Figure 5: Efficiency curve for benchmarking. Current utilization is located on the efficiency curve and related to the theoretical optimal efficiency, used for benchmarking the component.