From Code-Level Change to CI/CD Feedback: Empirical Validation of an Energy-Aware Software Framework

Abstract Purpose :: Energy stays invisible in the development workflow: continuous integration pipelines report build status, tests, and latency, but not that a change raised energy use. This study provides an initial empirical operational validation of a previously proposed energy-aware feedback loop that connects code-level intervention, CI/CD energy measurement, and energy-regression detection, testing whether these three technical components operate together on real software rather than proposing a new framework. Methods :: The framework was applied to two open-source Spring Boot applications and a GitHub Actions pipeline, using controlled before-and-after measurements at ten runs per condition and injected energy regressions. Differences were reported as deltas, with effect sizes and average power used to separate runtime from power effects. The pipeline classified the per-build and per-endpoint energies against a runner-calibrated baseline. Results :: The investigated code-level interventions produced application-specific differences in measured energy and execution duration. PetClinic showed an approximately 5–8\% reduction in total energy following the removal of an identifiable redundant repository lookup, whereas the RealWorld post-refactoring condition showed 21.1\% lower measured energy and 17.3\% shorter execution time. In both cases, the energy differences were primarily driven by the runtime rather than statistically established reductions in the average power. Within CI/CD evaluation, endpoint-level measurements showed low run-to-run variation in the evaluated baseline, and the pipeline detected and localized the injected regressions while distinguishing them from the observed runner variability. Conclusion :: The results provide an initial operational validation that code-level energy assessment, CI/CD energy feedback, and regression detection can function as an integrated feedback loop in the evaluated software systems. The results are bounded by the investigated interventions, applications, runner configuration, and injected regression scenarios; practitioner adoption remains to be evaluated separately.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-10739499/v1
Primary Topic
Green IT and Sustainability
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

From Code-Level Change to CI/CD Feedback: Empirical Validation of an Energy-Aware Software Framework

Muhammad Asif Khan, Jari Porras, Shola Oyedeji, Mbanyick Joof
Research Square
Green IT and Sustainability
preprint

From Code-Level Change to CI/CD Feedback: Empirical Validation of an Energy-Aware Software Framework

Muhammad Asif Khan, Jari Porras, Shola Oyedeji, Mbanyick Joof
preprint en

Abstract

Abstract Purpose :: Energy stays invisible in the development workflow: continuous integration pipelines report build status, tests, and latency, but not that a change raised energy use. This study provides an initial empirical operational validation of a previously proposed energy-aware feedback loop that connects code-level intervention, CI/CD energy measurement, and energy-regression detection, testing whether these three technical components operate together on real software rather than proposing a new framework. Methods :: The framework was applied to two open-source Spring Boot applications and a GitHub Actions pipeline, using controlled before-and-after measurements at ten runs per condition and injected energy regressions. Differences were reported as deltas, with effect sizes and average power used to separate runtime from power effects. The pipeline classified the per-build and per-endpoint energies against a runner-calibrated baseline. Results :: The investigated code-level interventions produced application-specific differences in measured energy and execution duration. PetClinic showed an approximately 5–8\% reduction in total energy following the removal of an identifiable redundant repository lookup, whereas the RealWorld post-refactoring condition showed 21.1\% lower measured energy and 17.3\% shorter execution time. In both cases, the energy differences were primarily driven by the runtime rather than statistically established reductions in the average power. Within CI/CD evaluation, endpoint-level measurements showed low run-to-run variation in the evaluated baseline, and the pipeline detected and localized the injected regressions while distinguishing them from the observed runner variability. Conclusion :: The results provide an initial operational validation that code-level energy assessment, CI/CD energy feedback, and regression detection can function as an integrated feedback loop in the evaluated software systems. The results are bounded by the investigated interventions, applications, runner configuration, and injected regression scenarios; practitioner adoption remains to be evaluated separately.

Research Square
Lappeenranta-Lahti University of Technology (FI)
Affordable and clean energy, Industry, innovation and infrastructure, Climate action
Green IT and Sustainability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon