Understanding and Predicting Docker Build Duration: An Empirical Study of Containerized Workflow of OSS Projects

Artifact package for the ASE'22 paper "Understanding and Predicting Docker Build Duration: An Empirical Study of Containerized Workflow of OSS Projects". This reproducible repository includes our data, scripts, and survey questions.

Abstract of the Paper

Docker building is a critical component of containerized workflow, which automates the process by which sources are packaged and transformed into container images. If not run properly, Docker builds can bring long durations (i.e., slow builds), which increases the cost in human and computing resources, and thus inevitably affect the software development. However, the current status and remedy for the duration cost in Docker builds remain unclear and need an in-depth study. To fill this gap, this paper provides the first empirical investigation on 171,439 Docker builds from 5,833 open source software (OSS) projects. Starting with an exploratory study, the Docker build durations can be characterized in real-world projects, and the developers’ perceptions of slow builds are obtained via a comprehensive survey. Driven by the results of our exploratory study, we propose a prediction modeling of Docker build duration, leveraging 27 handcrafted features from build-related context and configuration and 8 regression algorithms for the prediction task. Our results demonstrate that Random Forest provides the superior performance with a Spearman’s correlation of 0.781, outperforming the baseline random model by 82.9% in RMSE, 90.6% in MAE, and 94.4% in MAPE, respectively. The implications of this study will facilitate research and assist practitioners in improving the Docker build process.

Overview of the Docker Containerized Workflow

Citation

BibTeX

@inproceedings{WuASE22,
author={Wu, Yiwen and Zhang, Yang and Wu, Kele and Wang, Tao and Wang, Huaimin},
title={Understanding and Predicting Docker Build Duration: An Empirical Study of Containerized Workflow of OSS Projects},
booktitle={Proceedings of the 37th IEEE/ACM International Conference on Automated Software Engineering (ASE)},
year={2022},
organization={ACM}
}

Data

Data for prediction models
- We start with the dataset proposed by Lin et al.'s icsme20-Docker-dataset
- The data for prediction models can be found in models_data_cv.csv
- This data includes 171,439 builds with 27 features
Other data: builds, commits, and dockerfiles from 5,833 projects.
- 171,439 builds in builds.rar
- 551,838 commits in commits.part1.rar and commits.part2.rar
- 65,026 dockerfiles in dockerfiles.rar

Scripts

We implement the prediction models using R with the caret package.

Implementation details of prediction models: see prediction_models.md
Scripts for prediction modeling: see scripts_for_modelling.R
Optimal parameters settings of prediction models: see optimal_parameters.md
Methods used in our feature analysis
- VARCLUS clustering analysis (varclus in Hmisc package)
- Redundancy analysis (redun in rms package)
- Permutation importance (varImp in caret package)
Statistical hypothesis tests used in our quantitative analysis
- Wilcoxon signedrank test
- Cliff's delta
- ScottKnott ESD test
- Kruskal-Wallis test

Survey questions

Our survey questions can be found in survey_questions.md