Green Software Engineering in Practice: Quantifying the Impact of Carbon Reduction Strategies in Enterprise Software

Context. Measuring and optimizing software's carbon footprint is methodologically challenging. While the Software Carbon Intensity (SCI) specification standardizes carbon measurement, most research emphasizes infrastructure-over application-level optimizations. Goal. We aim to evaluate carbon emission reduction strategies for an industrial software application by applying the SCI specification. Method. This study targets AFAS SB, an enterprise application developed in the Netherlands. Firstly, baseline carbon emissions are assessed via the SCI specification from Microsoft Azure metrics collected in production. Then, in-The-lab experiments are conducted for quantifying how 18 application-level configurations impact the estimated CO2 emissions of AFAS SB across 79 scenarios. Results. We provide a concrete structured approach for applying the SCI specification to evaluate carbon emissions in enterprise applications. Garbage collection optimization reduced CO2 emissions by 29.5-30.6%, while other configurations had minimal effect. Combining top parameters achieved a 25.7% reduction without compromising functionality. Correlation analysis linked disk I/O reads to higher emissions. Conclusions. Our findings show that developers can reduce CO2 emissions through application-level configuration changes, with SCI enabling standardized measurement of optimization effectiveness across scenarios.