Flexecutor: Out-of-the-Box Smart Provisioning for Serverless Workflows

Serverless computing enables applications to scale dynamically without manual server management, with users paying only for the actual compute resources consumed. In the context of parallel workload execution, the problem of determining how many resources to provision and of what size has been extensively studied in recent years. However, existing approaches often present usability challenges, such as ad-hoc implementations or tight integration with a single cloud provider, which hinder the efficient deployment of new applications. In this work, we introduce Flexecutor, a Python library built on top of Lithops that unifies state-of-the-art serverless provisioners into a reusable and easy-to-use framework. Flexecutor abstracts away the underlying complexity, providing practitioners with a simple interface to deploy, manage, and optimize serverless workflows across multiple FaaS platforms. With Flexecutor, the implementation effort for coding serverless workflows is significantly reduced, while enabling the seamless integration of new smart provisioning strategies.