The path to serverless migration : an empirical investigation of decision-making process

Abstract

Serverless computing has redefined cloud computing by enabling organizations to focus exclusively on application development while abstracting away infrastructure management. This paradigm offers enhanced scalability, efficiency, and reduced operational overhead, making it well-suited for greenfield development and legacy system modernization. However, migrating to serverless architecture requires substantial system refactoring and demands careful alignment across technical and organizational dimensions, as it is inherently a socio-technical process. Furthermore, organizations must evaluate and select a suitable serverless platform based on their requirements to ensure the successful deployment of their applications.

To this end, the dissertation employs a mixed-methods approach, combining qualitative methods (interviews and case studies) with quantitative methods (repository mining and literature review) across five distinct publications. Through these publications, the dissertation explores the decision-making journey of migrating legacy applications to serverless computing and develops a decision model to support organizations in selecting the most appropriate serverless platform.

The findings reveal that organizations primarily migrate their applications to enhance scalability and reduce operational overhead. They adopt strategies such as domain-driven design and the strangler pattern to manage migration effectively. Despite these strategies, challenges persist, particularly in testing serverless applications and shifting development mindsets. Additionally, this dissertation frames the decision-making process of migrating applications to serverless architecture by identifying 29 key decisions and alternatives across technical and organizational dimensions. It further analyzes 12 prominent opensource serverless frameworks and presents a taxonomy of 263 types of issues and their 158 types of underlying causes to help organizations identify issue types and how to mitigate them. Finally, the dissertation introduces a decision model for selecting a feasible serverless platform, which is validated through five case studies. The case study participants confirmed that the model provided valuable insights, streamlined the selection process, and reduced both decision-making time and costs.

This dissertation advances both theoretical and practical understanding by decomposing the complexity of serverless migration. It provides structured guidance to help organizations navigate the technical, economic, and organizational challenges of adopting serverless computing, thereby improving their decision-making capabilities in this evolving field.