PPTAM$η$: Energy Aware CI/CD Pipeline for Container Based Applications

TL;DR

The paper addresses the gap that energy consumption is rarely measured in CI/CD for container-based microservices. It proposes PPTAMη, an automated, energy-aware pipeline that integrates load generation, container monitoring, and hardware power measurement into GitLab CI, enabling per-commit energy metrics and longitudinal analyses. The approach builds on PPTAM with energy measurement plugins and demonstrates feasibility via a JWT-authenticated containerized API across four commits, revealing how cryptographic algorithms and payload sizes affect energy usage while throughput remains stable. Findings show meaningful energy regressions can be detected early through integrated visualization and regression analysis, facilitating energy-conscious design choices and long-term research on energy efficiency in software. The work has practical significance for DevOps practitioners, testers, and researchers by providing a reproducible workflow for energy-aware regression testing in containerized environments.

Abstract

Modern container-based microservices evolve through rapid deployment cycles, but CI/CD pipelines still rarely measure energy consumption, even though prior work shows that design patterns, code smells and refactorings affect energy efficiency. We present PPTAM$η$, an automated pipeline that integrates power and energy measurement into GitLab CI for containerised API systems, coordinating load generation, container monitoring and hardware power probes to collect comparable metrics at each commit. The pipeline makes energy visible to developers, supports version comparison for test engineers and enables trend analysis for researchers. We evaluate PPTAM$η$ on a JWT-authenticated API across four commits, collecting performance and energy metrics and summarising the architecture, measurement methodology and validation.

Figures (3)

• Figure 1: Sequence diagram of pipeline process (BPMN notation).
• Figure 2: Diagram of PPTAM$\eta$ (C4 notation brown2026). blue indicate PPTAM$\eta$ components; green indicate used elements; orange indicate software under test;
• Figure 3: Response time, cpu utilizzation and power consumption of each version tested