Exploring Descriptive Metrics of Build Performance

Exploring Descriptive Metrics of Build Performance: A Study of GitHub Actions in Continuous Integration Projects

Constantinescu, Radu (TU Delft Electrical Engineering, Mathematics and Computer Science)

Proksch, S. (mentor)
Huang, S. (mentor)
Aivaloglou, E.A. (graduation committee)

Delft University of Technology

Computer Science and Engineering

CSE3000 Research Project

2023-06-28

The Continuous Integration (CI) practice, has been rapidly growing and developing ever since it's introduction. This practice has been constantly providing benefits to developers such as early bug detection and feedback to development teams. In this study, we aim to identify the descriptive metrics that best illustrate the performance of the CI build stage, regarded as heart of the development process.

We conduct a small case study on repositories utilizing GitHub Actions, a CI tool that is relatively unexplored. Within this context, we classify projects using two performance indicators: build breakages and build durations. We examine two distinct sets of metrics in our analysis. The first set being build level metrics, which are closely linked to the build stage. The second set including project level metrics.

Our findings suggest that patterns traditionally associated with low breakages and durations are applicable to repositories employing GitHub Actions. However, understanding the relationship between project level metrics demands a more comprehensive approach, necessitating a thorough analysis of the project context for a holistic understanding of build performance.

Build Performance
Continuous Integration
GitHub Actions
Open source software

http://resolver.tudelft.nl/uuid:2bbd8857-3dc0-4e52-8f6a-6518c557d287

Student theses

bachelor thesis

© 2023 Radu Constantinescu