Analysis and reengineering of a cloud-based distributed system responsible for the CI/CD processes of the Falco open-source project.

Abstract

In recent years, CI/CD architectures have drastically revolutionized the collaborative software development cycle approach, allowing for a massive improvement resulting from the automation of builds, testing, integration, and deployment pipelines. Nowadays, CI/CD solutions often leverage third parties cloud-based distributed systems to accomplish their tasks. Organizations pay cloud providers’ services on a per-usage basis, and this is a motivation to put relevant efforts into minimizing costs and maximizing the performance and efficiency of these CI/CD systems. This field gets even more interesting in open-source scenarios, where contributions from different parties must be taken into account and managed with attention to keeping processes and workflows transparent to the community. This thesis explores the Falco project from this perspective: Falco is open-source software that will be used as a case study to break down its current CI/CD system. The work done on this project includes the collection of qualitative and quantitative analyses (related to its design and performance, respectively), along with proposals for possible improvements and their implementation. This final step will lead to a reengineering of the cloud architecture, resulting in an enhanced way to manage pipelines and workflows, with particular attention to the ones responsible for dealing with the organization’s secrets. The thesis then compares the current CI/CD design with other possible strategies, relying on different state-of-the-art tools than the ones adopted. Moreover, the comparison will span to the old implementation, pointing out advantages and disadvantages, also concerning the differences in terms of performance. The performance comparison is possible thanks to data and metrics collection done before and after the reengineering.