Quelltext der Seite Performance Modeling and Evaluation of an MQTT–Kafka Based Data Streaming Architecture

{{Vortrag
|vortragender=Bescher Kilani
|email=ufsrc@student.kit.edu
|vortragssprache=Deutsch
|vortragstyp=Bachelorarbeit
|betreuer=Maximilian Hummel
|termin=Institutsseminar/2026-04-17
|vortragsmodus=in Präsenz
|kurzfassung=This thesis presents an architectural performance modeling approach for an IoT data streaming pipeline integrating MQTT and Apache Kafka. The goal is to assess to what extent performance modeling can forecast throughput, latency, and resource utilization under varying workloads. Using the Palladio Component Model (PCM), a component-based architectural performance model (APM) of an open-source MQTT-to-Kafka pipeline is developed. Resource demands are systematically extracted via profiling and transformed into stochastic model parameters. The calibrated model is validated against empirical measurements of the running system, achieving prediction accuracies of up to 99% at calibration points, with decreasing accuracy between workload intensities. Finally, the model is used to extrapolate performance behavior beyond the measured baseline. The results demonstrate that PCM-based architectural modeling can represent continuous, message-driven IoT workloads and provide meaningful performance forecasts before deployment.
}}