Replication package and Online Appendix

This is a replication package and online appendix for the ICSA2025 paper "Network Centrality as a New Perspective on Microservice Architecture"

Contents

This repository contains the following:

• INSTALL: Detailed installation instructions for each used tool
• Appendix:
  • Table I: Data on the 24 studied projects
  • Table II: List of all gathered metrics
  • Project information: Tables of all projects gathered from all dataset and their metadata
• Projects: a folder containing source code of all studied projects
• Figures: a folder containing all figures used in the paper
  • Sankey diagram: Sankey diagram of project selection process
  • BMP process: BMP process for the data collection
  • Size Heat Map: Heat of Spearman Rho for correlation of size metrics with centrality
  • Size Heat Map (Stat. Sig.): Heat of Spearman Rho for correlation of size metrics with centrality (stat. sig. only)
  • Complexity Heat Map: Heat of Spearman Rho for correlation of complexity metrics with centrality
  • Complexity Heat Map (Stat. Sig.): Heat of Spearman Rho for correlation of complexity metrics with centrality (stat. sig. only)
  • Quality Heat Map: Heat of Spearman Rho for correlation of quality metrics with centrality
  • Quality Heat Map (Stat. Sig.): Heat of Spearman Rho for correlation of quality metrics with centrality (stat. sig. only)
  • Centrality Heat Map (Stat. Sig.): Heat of Spearman Rho for correlation of centrality metrics with each other (stat. sig. only)
• Raw data: a folder containing all raw data extracted from different tools
  • code2dfd: Raw output of Code2DFD
  • graph: Graphs extracted from Code2DFD output
  • understand: Raw data from Understand
  • jasome: Raw data from Jasome
  • package_map.json: Mapping of Java packages to microservices
• Metrics: metrics extracted from raw data
  • metrics_centrality.csv: All the centrality metrics for all microservice
  • metrics_understand.csv: All the Understand metrics for all microservices
  • metrics_jasome_package.csv: All the Jasome metrics for all microservices on package level
  • metrics_jasome_class.csv: All the Jasome metrics for all microservices on class level
  • metrics_jasome_method.csv: All the Jasome metrics for all microservices on method level
  • metrics_sonarqube.csv: All the SonarQube metrics for all microservices
  • metrics_merged.csv: All the metrics for all microservices
  • metrics_statsig.csv: List of all metrics that have a statistically significant correlation with centrality
• Results: Data files containing the analyzed results to answer the Research Question
  • NormalityAndersonDarling: Results of testing normality of each metric distribution with Anderson-Darling
  • RQ1: Does centrality correlate with size metrics?
    • metrics_size.csv: All the size metrics that have a statistically significant correlation with centrality
    • SizeSpearmanRho: Spearman Rho correlation between centrality and size metrics
  • RQ2: Does centrality correlate with complexity metrics?
    • metrics_complexity.csv: All the complexity metrics that have a statistically significant correlation with centrality
    • ComplexitySpearmanRho: Spearman Rho correlation between centrality and complexity metrics
  • RQ3: Does centrality correlate with quality metrics?
    • metrics_quality.csv: All the quality metrics that have a statistically significant correlation with centrality
    • QualitySpearmanRho: Spearman Rho correlation between centrality and complexity metrics
• Scripts: a folder containing all the scripts
  • extract_graphs.py: processes the Code2DFD output into standard graph json
  • metrics_centrality.py: computes centrality scores with NetworkX
  • metrics_jasome.py: executes Jasome tool and saves raw data
  • merge_jasome.py: merge the raw Jasome data into csv files
  • metrics_understand.py: executes Understand tool and saves raw data
  • merge_understand.py: merge the raw Understand data into a csv file
  • metrics_sonarqube.py: executes SonarQube analysis
  • merge_sonarqube.py: put the SonarQube data into a csv file
  • merge_all.py: merge all metrics into a single csv file
  • metrics_results.py: keep only the statistically significantly correlated metrics

License

All generated data is provided under Creative Commons 4.0 Attribution License.

All scripts are provided under the MIT License.

All the analysed projects must be used in accordance with their respective licenses (shared in each project when applicable).

Preparation and installation

Follow the instructions in INSTALL to install and configure all used tools.

Code2DFD and centrality metrics

All DFDs are reconstructed with c65b4a version of Code2DFD tool.

The raw output for each PROJECT is copied here to a folder PROJECT-code2dfd.

Converting Code2DFD to NetworkX graph

The script process_graphs.py converts the json files of the Code2DFD output into network files suitable for NetworkX.

For each PROJECT, it creates 2 files in a PROJECT-graph folder:

• PROJECT-gwcc.json: The Greatest Weakly Connected Component (GWCC) of the reconstructed architecture graph
• PROJECT-gwcc_noDB.json: The GWCC with all databases that are only connected to one service removed

Computing centrality metrics

The script metrics_centrality.py loads the PROJECT-gwcc_noDB.json files and computes the following centrality metrics with NetworkX:

• Degree centrality
• In-degree centrality
• Out-degree centrality
• Eigenvector centrality
• Closeness centrality
• Betweenness centrality
• Load centrality
• Harmonic centrality
• Information Centrality
• Current flow centrality
• Subgraph centrality

The centrality metrics for each system and service are saved into metrics_centrality.csv csv file.

Jasome metrics

Jasome tool can be downloaded from its GitHub page.

Raw Jasome data

The script metrics_jasome.py executes the Jasome tool for each PROJECT.

Change the variable JASOME_PATH in the script to point to the Jasome binary on your system.

For each PROJECT, the scripts saves to the folder PROJECT-jasome the raw xml output from Jasome for each src folder in the project.

Merging Jasome metrics

The script merge_jasome.py takes the data from all the raw xmls into the following csv files:

• metrics_jasome_package.csv: metrics calculated for each package
• metrics_jasome_class.csv: metrics calculated for each class
• metrics_jasome_method.csv: metrics calculated for each method

Understand metrics

Download the Understand tool and acquire its license on the official website

Raw Understand data

The script metrics_understand.py executes the Understand tool for each PROJECT.

Change the variable UND_PATH in the script to point to the und cli tool on your system.

For each PROJECT, the scripts saves to the folder PROJECT-und the raw csv output from Understand.

Merging Understand metrics

The script merge_understand.py takes only the metrics calculated on Package level for all PROJECTS and saves them to metrics_understand.csv csv file.

SonarQube metrics

Deploy a SonarQube instance using the instructions from the official website.

Generate a Global Analysis Token and a User token.

Download the SonarScanner application from the official website.

Raw SonarQube data

The script metrics_sonarqube.py sets up a SonarQube project for each of the PROJECTs in the repository and executes the analysis with SonarScanner.

Change the SONAR_PATH variable to the location of the sonar-scanner binary.

Change the TOKEN variable to the Global Analysis Token generated in SonarQube.

Additionally, if SonarQube is not deployed on localhost:9000, change the -Dsonar.host.url parameter in the run command.

After executing the script, you should see all projects analyzed in the SonarQube dashboard.

Merging SonarQube data

The script merge_sonarqube.py queries data for each PROJECT.

Change the variable USER_TOKEN to the User token generated in SonarQube.

The script will query the SonarQube metrics on directory level, infer the package name from the directory path, and save the metrics for each PROJECT and each package in the metrics_sonarqube.csv.

Merging all data

The file package_map.json contains the mapping of Java packages to the microservices.

The script merge_data.py takes the metrics_centrality.csv, metrics_understand.csv, metrics_jasome_package.csv, metrics_jasome_class.csv, metrics_jasome_method.csv , metrics_sonarqube.csv files, maps the packages to microservices and creates a unified csv file metrics_merged.csv with microservices that have all possible metrics.

Metrics are aggregated from packages using sum, mean and max wherever suitable.

Filtering metrics with statistically significant correlation

The file metrics_statsig.csv contains a list of metrics that have a statistically significant (p<0.01) correlation with at least one centrality score, and their category of either size, complexity or quality.

The script metrics_filter_statsig.py takes only such metrics from metrics_merged.csv and saves them to three respective csv files: metrics_size.csv, metrics_complexity.csv and metrics_quality.csv together with the centrality metrics.