Edge Computing on Embedded Systems using WebAssembly

This thesis explores the viability and performance of WebAssembly (Wasm) for edge computing on embedded systems. Specifically, it evaluates whether WebAssembly, when executed on a lightweight embedded operating system such as Zephyr, can deliver adequate computational efficiency for typical edge computing scenarios. To assess this, a project was built using WebAssembly Micro Runtime (WAMR) integrated in Zephyr. A benchmarking program was developed to test common edge computing performance indicators and tasks such as Fast Fourier Transform (FFT) and Matrix Multiplication under varying code optimization levels and runtime configurations. Results show that WebAssembly introduces significant overhead compared to native code execution when running in interpreter modes. However, it provides benefits with respect to portability, security, and sandboxing. The findings suggest that WebAssembly running in interpreter modes is not suitable for compute-intensive edge workloads on the i.MX RT1180 microcontroller because of the execution time overhead introduced and the lack of support for the WebAssembly System Interface (WASI) standard.