Decreasing Build Time in Open Build Service Through Containerization: Implementation and Performance Analysis

Abstract

The Open Build Service (OBS) enables reproducible software package creation but suffers from a performance bottleneck during build initialization. Each build requires extracting compressed archive files containing base systems and dependencies. When multiple builds share common dependencies, this repeated extraction of identical content creates significant time overhead, particularly in enterprise environments with frequent builds. This thesis analyzes the OBS architecture and implements containerization to replace traditional archive extraction. The implementation modifies three core components: client enhancements for preinstallimage.info file downloads, build script containerization logic with hash-based image tracking, and worker integration supporting Podman alongside existing virtualization technologies. Performance evaluation compares Docker and Podman container engines using preinstall images as container base images. Podman containerization delivers build time reductions ranging from 8 to 90 seconds depending on dependency complexity, translating to substantial time savings in enterprise environments executing hundreds of daily builds. Most significantly, containerization enables Podman to utilize preinstall images in rootless environments for the first time, transforming it from the slowest traditional approach to achieving optimal 5-second build times. Both engines demonstrate network efficiency improvements of approximately 45%. The containerization approach provides faster builds while maintaining backward compatibility with OBS infrastructure and security requirements.