Abstract
Function-as-a-Service (FaaS) receives close attention due to highly desirable characteristics, including pay-as-you-go pricing, high elasticity, and its fully managed nature. To leverage these benefits for existing applications, developers face the challenge of migrating legacy code to a FaaS platform (FaaSification). Unfortunately, however, actionable guidance on how to do so for real-world applications does not exist. In this paper, we report on our experience from FaaSifying a data-intensive application, and evaluating different options through extensive experimentation, using approaches such as regression tests and tracing. Based on the obtained results, we present five migration tactics in support of future FaaSification.