Serverless workflow execution models and engines

Serverless computing is a cloud service model where developers can build and run applications without managing servers or backend infrastructure. The serverless computing model assumes that during the application development, deployment and execution, there is no need to maintain any server infrastructure. In contrast, provisioning the infrastructure is left to the cloud provider, as in the Function-as-a-Service (FaaS) or Container-as-a-Service (CaaS) models. The serverless computing paradigm is appealing due to its abstraction of provisioning and infrastructure management complexities. Resources are dynamically scaled based on demand, and billing operates on a pay-as-you-go model, charging only for execution time. This eliminates costs associated with idle capacity. Serverless computing offers an interesting option for workflow execution, as it provides a transparent and dynamic way of provisioning resources for workflow tasks. Cloud functions in a FaaS model are stateless and independent, so a workflow model becomes a natural solution for their orchestration. In this chapter, we provide an overview of serverless computing applied to workflow applications, discuss the serverless workflow execution models, and survey the workflow execution engines supporting the serverless model. We provide a taxonomy of selected features of serverless workflow engines, including workflow standards, centralized vs distributed execution, deployment environment, data management, and scalability. We conclude with examples of applications and an outlook for future work.