How Accurately Can the Energy Use of Spark Applications Be Estimated Based on Resource Utilisation?
2026-08-10 • Distributed, Parallel, and Cluster Computing
Distributed, Parallel, and Cluster Computing
AI summaryⓘ
The authors studied how to accurately estimate the energy use of big data processing tasks running on cloud and local computer clusters. They focused on Apache Spark jobs managed by Kubernetes and compared energy estimates from models with actual hardware energy measurements using Intel's built-in counters. Their results showed that using external monitoring data improved the accuracy of energy estimates, reducing how much the models underestimated energy consumption on both cloud and local setups. This helps in better tracking the sustainability impact of distributed computing tasks.
energy estimationApache SparkKubernetesIntel RAPLcloud computingbatch data processingresource utilizationenergy monitoringbare-metal cloudsustainability accounting
Authors
Youssef Moawad, Kathleen West, Vasilis Bountris, Philipp Thamm, Yehia Elkhatib, Lauritz Thamsen
Abstract
Distributed batch data processing applications are widely executed on cloud-based resources where restricted user access to node-level hardware energy counters hinders transparent sustainability accounting. Energy and carbon attribution methodologies therefore depend on power models and available resource utilisation traces, yet the accuracy of these estimates has to be validated while direct counters are available. In this work, we use Apache Spark running on Kubernetes as a case-study dataflow runtime and cluster resource manager to compare model-based energy estimates to Intel RAPL package and DRAM energy on an AWS bare-metal cloud and an on-premises cluster, comparing different CPU usage signals and memory coefficients. We show that external monitoring improves signed package-energy error relative to Spark task traces, reducing underestimation from -29.58% to -24.41% on AWS and from -24.00% to -16.22% on-premises.