FUSPAQ: A Function Selection Platform to Adjust QoS in a FaaS Application

Function as a Service (FaaS) development has numerous benefits for application deployment, management, and maintenance. However, the lack of control over the infrastructure and, often, over the FaaS platform itself, makes it necessary to look for external solutions that allow the operation of the application to be adapted to different requirements or changing execution conditions. In a FaaS application, the quality of service (QoS) is determined by the characteristics of the functions executed to perform each workflow operation. Deciding the most suitable functions providing a QoS is a complex process due to the high variability of possible function implementations, each giving different qualities. Leaving this task in the hands of the developer is not a good solution and makes it difficult to program the application. We present FUSPAQ, a framework for working with serverless architectures, which can automatically select the best functions executed at runtime to satisfy specific QoS requirements. With this objective, a Software Product Line approach is used, modeling the application’s tasks and operations using Feature Models that specify the variability of functions that can perform the same operation as a family of functions. We use Z3, a cross-platform satisfiability modulo theories (SMT) solver, to generate optimal configurations. As requirements can change over time, the system automatically adapts to these changes to continue maintaining the desired QoS. We test our approach with different QoS parameters, and analyse the value added to serverless frameworks.