Replication package for the paper:

Introducing Interactions in Multi-Objective Optimization of Software Architectures

https://arxiv.org/abs/2308.15084

---

How to generate the tables and figures in the paper

Initialize the python execution environment:

git clone https://github.com/SpencerLabAQ/replication-package__interactive-search-based-software-performance.git
cd replication-package__interactive-search-based-software-performance
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

RQ1: quality of solutions

Compute the quality indicators and generate Table 4:

python rq1_quality.py

The previous command will also generate two CSV files (ttbs_metrics_by_run.csv and ccm_metrics_by_run.csv) which contain a run by run computation of the quality indicators, and are used to generate the plots of RQ1.

Generate the plots in Figures 4 and 5:

# Train Ticket case study
python rq1_quality_plots.py ttbs_metrics_by_run.csv ttbs

# CoCOME case study
python rq1_quality_plots.py ccm_metrics_by_run.csv ccm

The resulting plots will be saved in the plots folder.

RQ2: architectural differences

Execute notebooks rq2-cocome.ipynb or rq2-trainticket.ipynb to generate the charts for Figures 7 and 8.

Among other computations, the notebooks above will generate several CSV and DOT files under the datasets folder.

Execute notebook tree-scripts.ipynb to generate the charts for Figure 6.

The datasets needed by the notebooks are automatically loaded from a default folder.

RQ3: covered design and solution spaces

Execute notebooks rq3-cocome.ipynb or rq3-trainticket.ipynb to generate the charts for Figures 9 and 10.

Among other computations, the notebooks above will generate several CSV and DOT files under the datasets folder.

Execute notebook tree-scripts.ipynb to generate the charts for Figures 11 and 12.

The datasets needed by the notebooks are automatically loaded from a default folder.

Experiments

The experiments in the paper were performed using EASIER, which is available in a different repository:

https://github.com/SEALABQualityGroup/EASIER

All the data gathered during such experiments is provided here, in the zips folder.