Microservice systems are hard to understand: the behaviour of a single request is spread across many independently deployed services and visible nowhere in full. Distributed tracing frameworks such as OpenTelemetry record that behaviour one request at a time, but a pile of individual traces does not reveal the general patterns of interaction that characterise a system. Process mining can turn such data into structured models automatically, but when applied naively to OpenTelemetry traces, it results in unintelligible and incorrect models.
This thesis asks how process mining can be applied to OpenTelemetry traces so that the result helps engineers and architects understand, validate, and improve their systems. Its answer is layered process mining: rather than one global model, mine one model per service operation from only the direct children of that operation’s spans, and represent each call to a child operation as an expandable subprocess reference. This preserves the hierarchy of a trace end to end, producing models that are structurally faithful and stay navigable however large the system grows. The approach is grounded in interviews with six engineers and architects at Info Support and realised in Ariadne, an open-source framework that mines a hierarchical model per operation with a lifecycle-aware Inductive Miner, checks conformance against a reference architecture, and renders the models as an interactive BPMN visualisation.
Evaluated on the MediaMicroservices benchmark, the mined model of its most complex endpoint matches the application source code, capturing concurrency a completion-order miner would have serialised. Evaluated on Uber’s CRISP dataset of 100,000 production traces, Ariadne mines all 18,386 operations in under six minutes on a single workstation while keeping model quality high. Once the hierarchy of traces is preserved rather than discarded, process mining becomes a practical lens on runtime microservice behaviour. Keywords: process mining; distributed tracing; OpenTelemetry; microservice architecture; conformance checking; software observability.