A Prototype Implementation of SINETStream for Resource-Constrained MicroPython Environments

Abstract

SINETStream is a messaging library designed for wide-area IoT data collection. While its standard Python and Java implementations have been publicly available and actively used since 2019, primarily targeting personal computers and single-board computers, directly deploying these libraries onto microcontroller units (MCUs) is impractical due to severe memory and storage constraints. In this paper, we propose a MicroPython-based implementation of SINETStream optimized for resource-constrained edge devices. To minimize the memory footprint, we have developed a prototype focusing on MQTT. We addressed strict hardware limitations by replacing highoverhead dependencies with micro-libraries and transitioning configuration parsing from YAML to JSON, thereby reducing both operational overhead and the library's footprint. To maintain essential functional compatibility with the broader SINETStream ecosystem, this prototype retains the original modular architecture, supporting both Publisher and Subscriber roles. Furthermore, it ensures payload interoperability with existing data by incorporating Apache Avro serialization and gzip compression. Experimental results using a Raspberry Pi Pico 2 W demonstrate that the compiled library occupies only about 330 KB of flash memory. Even during continuous data transmission over Wi-Fi, total RAM consumption-including the MicroPython runtime environment-remained at approximately 200 KB. This efficiency reliably leaves more than half of the system memory available for user-defined sensing, data filtering, and actuator control logic. Ultimately, this implementation successfully extends the SINETStream ecosystem to the extreme edge using lowcost hardware. Given the widespread accessibility of inexpensive MCUs, we anticipate this lightweight adaptation will encourage broader utilization, including in educational applications.