Design of a distributed IoT system architecture for greenhouse microclimate monitoring based on MQTT protocol and edge computing

The article presents the design of a distributed IoT system architecture for greenhouse microclimate monitoring based on the MQTT protocol and edge computing. The relevance is driven by the low proportion of automated greenhouse facilities in the Russian Federation and the high deployment cost of traditional industrial controller-based systems. A four-layer architecture was developed, comprising a sensor node layer (ESP32-S3), an edge gateway layer (Raspberry Pi 4B), an MQTT broker layer (Mosquitto 2.0), and a cloud platform layer (PostgreSQL + Grafana). A comparative analysis of data transfer protocols (MQTT v5.0, CoAP, HTTP/2, AMQP 1.0) was conducted based on latency, throughput, traffic consumption, and power consumption criteria. MQTT v5.0 provides an optimal balance of throughput (12,400 msg/s) and latency (47 ms) with minimal traffic volume. An original methodology for calculating an integral system performance index accounting for latency, throughput, and message loss rate was proposed. Load testing on a testbed (50–400 sensor nodes, 72 hours) demonstrated that a configuration with 4 edge gateways ensures stable operation with 400 nodes at 89 ms latency and losses below 0.05%. The integral performance index reached 0.91 for the optimal configuration. It is shown that the application of edge computing reduces traffic volume between the facility and the cloud by 78% through local data aggregation.