This thesis presents a comprehensive simulation and analysis framework that examines how traffic characteristics emerge within different routing environments. The study employs Mininet to construct 5 different topologies and generate controlled traffic using 7 different protocols. A seven-dimensional feature vector was extracted from packet streams, capturing essential statistical properties including distribution shape, variability and temporal behaviour. To study how communication patters evolve, an overlapping sliding-window method was applied, producing high resolution temporal snapshots. Each window was compared to a global feature vector using cosine distances, enabling the measurement of deviations over time. The resulting feature was analysed using DBSCAN, for its ability to detect arbitrarily shaped clusters and isolate anomalous traffic without requiring a predefined number of clusters.