Serverless computing has revolutionized cloud-based software development for software developers, addressing many of the associated challenges. With resource management and infrastructure provisioning handled by the provider, developers can focus on deploying services at the application level, which has gained significant popularity. Edge computing, with its proximity to end-users and ability to offer low-cost, high-speed services, has emerged as a promising solution for latency-sensitive applications. By providing Function-as-a-Service (FaaS) capabilities at the network edge, it becomes easier to leverage the available edge capacity. However, edge servers have limited capacity to provide FaaS services, and user access is typically restricted to their geographic region. This presents a challenge in achieving optimal resource allocation in a multi-user environment, given the dynamic and unpredictable nature of edge computing. To address these challenges, we have proposed a dynamic resource provisioning and allocation approach for Edge FaaS. Our approach models the application structure as a directed acyclic graph, representing a workflow of computational functions. Drawing inspiration from the many-to-many assignment mechanism used in the course allocation problem, we have developed a mechanism to assign multi-user workflows to the limited Edge FaaS resources. This mechanism considers the priority of users’ ready tasks, determined by upward rank, with the goal of reducing workflow execution time. Additionally, we have proposed two methods for on-demand provisioning: the Highest Bid First Mechanism (HBFM) and the Warm Function First Mechanism (WFFM). Experimental results have shown that the proposed resource allocation methods effectively allocate limited resources to users, leading to reduced workflow completion time.

Keywords