PERFORMANCE METRICS FOR IOT-BASED SMART LEARNING SYSTEMS: A QUASI-EXPERIMENTAL ANALYSIS OF SCALABILITY AND LATENCY IN MULTI-CLASSROOM DEPLOYMENTS IN HYDERABAD

Abstract

Internet of Things (IoT) technologies that have proliferated in higher education have triggered the creation of smart learning spaces combining sensor networks, real-time analytics, and adaptive systems. These deployments have not however been studied in terms of technical scalability, especially in relation to network protocol performance with different device densities. The present study is a quasi-experimental analysis of the impact of the simultaneous loading of a system of IoT-based smart learning systems in 15 classrooms of one university in Hyderabad during a single academic semester on the measures of performance of the systems regarding their load capacity, frequency of data transmission, and communication protocols (MQTT vs. CoAP). We manipulated the independent variables (50, 100, and 200 IoT devices per classroom; 1 Hz, 5 Hz and 10 Hz transmission rate) to get dependent variables such as end-to-end latency, packet loss rate, energy consumption, and perceived responsiveness by the user. They used network analyzers and power monitoring equipment to collect data, and then they analyzed the results using repeated-measures ANOVA and multiple regression modeling. Findings (n = 2,160 observation intervals) reveal considerable main effects (p <.001) of the number of devices on latency ( 2154) = 147.32) and packet loss (, 2154) = 89.76), where MQTT has a 23.4% lower latency, and CoAP has 18.7% higher power consumption than the other at full load. Regression analysis attributed 73.8% of the variation in latency to the model ( =.738, RMSE = 12.4 ms) and found a scalability limit at 180 devices at which point the latency was too large to be useful in pedagogy (> 200 ms). The use of time-series analysis showed the patterns of diurnal performance degradation with congestion of the campus network. Results can serve as empirical reference points to the IT administrators of higher education institutions and offer a tested performance measurement framework to IoT deployments in learning settings.