An approach for modeling the operational requirements of FaaS applications for optimal deployment

FaaS can extend cloud capabilities to local edge devices. They enable composing applications into workflows and distributing their processes between the edge and the cloud. When deploying a FaaS application, the key issue is to define the desired operational requirements and find a configuration that satisfies them. Both problems are complex and subjective as they differ from one application to another depending on the application structure, run-time model, and optimization requirements. We address these issues using a general multi-criteria optimization approach based on a fuzzy analytical hierarchy process (AHP). Firstly, the specialists intuitively specify their fuzzy privacy in terms of data locality, cost, and performance requirements. Then, a Fuzzy AHP model is constructed to compare and select the optimal workflow configuration that satisfies the requirements. The work is evaluated using a real FaaS application, where AWS Cloud and AWS Greengrass present the cloud and the edge. We assessed the ability of the proposed approach to retrieve the optimal configurations for various scenarios and compared the results to one of the state-of-the-art approaches. Unlike existing hard-coded approaches, our approach is intuitive, modular, extensible, and addresses more DevOps requirements.