Description

This thesis explores the application of process mining techniques for analyzing the runtime behavior of distributed microservice-based systems. ProMiSE, a trace analysis toolbox, is developed to collect, transform, and evaluate service-level execution data. Traces are captured using OpenTelemetry and Jaeger, converted into structured event logs, and analyzed using process discovery and conformance checking algorithms. The toolbox supports the generation of process models in Petri net and BPMN form and provides additional insights through trace clustering, performance metrics, and visualizations such as Gantt charts and dotted diagrams. In addition, the approach enables a preliminary assessment of architectural conformance by comparing discovered runtime behavior with expected interaction patterns. While it does not cover all aspects of architectural correctness, it can reveal behavioral deviations and inconsistencies that may hint at architectural misalignments. Evaluation is conducted using both a controlled mock service setup and a real-world test system, focusing on the interpretability and diagnostic potential of the discovered models. The results demonstrate that behavioral insights can be extracted from trace data, even in low-variability environments. While the real-world evaluation is limited by trace diversity and observation period, the findings indicate that process mining can serve as a valuable complement to traditional monitoring in understanding and reasoning about software execution and aspects of architectural behavior at runtime.