Simulation Engine
- MAC-learning switch — each switch pre-seeds its forwarding table from the topology wiring, eliminating spurious flooding toward silent endpoints (e.g. a server that only receives). Dynamic learning still operates for traffic through intermediate switches. With
mac_table="topology"a switch instead starts with the port towards every end station (shortest path through switches, ties to the lowest port), i.e. already converged, and never floods unicast. - NIC filtering — end stations accept only frames addressed to their own MAC; flooded frames for other stations are dropped and do not count towards metrics.
- Store-and-forward model — transmission delay + propagation delay per hop.
- Queuing — each link direction has its own egress queue with 8 strict 802.1Q priority classes (FIFO within a class, non-preemptive);
qos="fifo"on a switch ignores priorities.queue_depthreports buffered frames at the switch's outbound ports. - Traffic shaping — constant, Poisson, and Pareto-burst patterns; IMIX frame-size distribution.
- Live mode — simulation runs in 200 chunks (~8 s total wall time); the dashboard reads shared metric arrays between chunks via Python's GIL.
The engine runs a Scenario, the execution-independent description that Network.simulate() builds from the network. It never reads the Network builder. It is the reference implementation of the Model Semantics. It is powered by SimPy — every node type (Adapter, Switch, MQTTBroker, ModbusSlave, ProfinetDevice) runs as one or more SimPy processes, and links are modelled as directed channels with their own transmission and propagation delay.