ESP32-based devices are widely used as a low-cost and resource-constrained platform for Internet of Things (IoT) applications that require responsive communication. However, real-time chatbot systems need higher requirements regarding delays that can reduce the perception of responsiveness, conversation continuity, and user trust. This study evaluates the use of WebSocket, Hybrid MQTT-WebSocket, Hybrid MQTT-Server-Sent Events (SSE), and Cloudflare as four real-time communication architectures for an ESP32-based chatbot system. This study used a systematic literature review on ESP32 communication models and experimental performance evaluation using Sovi Lau 2.0 logs. This dataset consists of 400 communication logs from the performance of the four architectures. The four metrics analyzed were latency, throughput, packet loss, and power consumption. One-way ANOVA was used to compare the architectures, followed by Tukey's HSD for significant metrics. The results show that Cloudflare has the lowest average latency and the highest throughput, while Hybrid MQTT-SSE produces the lowest packet loss and the lowest power consumption. The ANOVA test showed that of the variables tested, only latency had a significant difference between the architectures. The Tukey HSD test confirmed that the Cloudflare architecture had the significantly lowest latency.
Copyrights © 2026