Empirical evidence on technical challenges when adopting continuous practices

• André Portela UNICAMP
• Breno Bernard Nicolau de França UNICAMP

Resumo

Context: Continuous Software Engineering offers several benefits like fast iterations, increased reliability, and shorter lead times to react to business needs through practices like Continuous Integration, Delivery, and Deployment, but these practices come with many adoption challenges, which may appear as social (hard to identify and overcome) or technical challenges (TCs). Since anticipating even a small amount of possible challenges can be overwhelming, there is value in detecting them as early as possible to make it cheaper, faster, and more feasible to adopt this set of practices. Objective: We summarize evidence about identifying TCs to adopt continuous deployment and related practices. Method: We performed a Rapid Review that analyzed 14 studies and performed a narrative summary to provide an evidence briefing to practitioners. Results: We identified 121 TCs and organized them into nine categories: architecture, building, code, deployment, development practices, operations, pipeline, security, and testing. TCs related to code, pipeline, and testing are the most prevalent in the reviewed literature, while operations and deployment are the least. Static analysis strategies are the most used for detecting TCs, accounting for 94 detection strategies out of 100. However, most of these static analyses resort to some parsing technique. Specifically, Abstract Syntax Tree is the most used, detecting 44 TCs. Conclusion: Most studies have mixed evidence about identifying TCs to adopt continuous practices - some focus on evaluating the efficiency of detection strategies. In contrast, others are more focused on their characterization. We also identified varying levels of efficiency when detecting TCs. Finally, we advise attention to interpreting the results based on the overall quality score.

Palavras-chave: rapid review, continuous software engineering, continuous integration, continuous deployment, continuous delivery