A Comparative Performance Study of Parallel MQTT and RS485 Communication Architectures for High-Frequency IoT Sensing

High-frequency data collection in large-scale IoT sensing systems requires communication architectures that can maintain low latency and stable throughput as sensor density increases. Conventional RS485-based polling structures suffer from rapid performance degradation under multi-node and high-rate conditions due to their sequential communication model. In this study, we present a comparative performance analysis of parallel MQTT and RS485 communication architectures for high-frequency IoT sensing. The proposed parallel MQTT structure is implemented with topic-level parallelism, QoS-based reliability control, and non-blocking scheduling, and its performance is quantitatively evaluated under multi-sensor experimental conditions. Experimental results show that the parallel MQTT architecture achieves lower average latency (98 ms vs. 178 ms), higher message success rate (99.2% vs. 94.5%), and higher throughput compared to the RS485-based system. In addition, we analyze message loss behavior under high-frequency operation and examine how topic hierarchy and asynchronous processing affect broker-side congestion.