Kilian Müller
Friedrich-Alexander-Universität Erlangen-Nürnberg
Maximilian Seidler
Friedrich-Alexander-Universität Erlangen-Nürnberg
Peter Ulbrich
TU Dortmund
Norman Franchi
Friedrich-Alexander-Universität Erlangen-Nürnberg

NimbleNet: Serverless Computing for the Extreme Edge in Factory Environments

Modern factories are moving towards modular layouts with more generic production cells, allowing rapid production customization. Here, the cell’s edge nodes exhibit considerable heterogeneity, ranging from simple microcontrollers to fully-fledged computing systems, which mandates that assembly-specific serverless functions used by Automated Guided Vehicles (AGVs) navigating through and linking the production cells are platform-independent. These moving manufacturing processes call for an adapted communication and distribution infrastructure to maintain low response times while minimizing network traffic. We present NimbleNet, a lightweight distribution and orchestration approach for serverless functions specifically tailored for IoT devices in industrial factory settings. NimbleNet leverages lightweight sandboxing such as WebAssembly, dynamic dependency management, and a neighbor-first caching strategy to enable efficient, platform-independent deployment of tasks at the extreme edge. Our simulations and real-world evaluations demonstrate that our approach facilitates the dynamic deployment and execution of tasks on resource-constrained devices while optimizing the network load within a mesh network configuration.