Extension points: the eight Simulation setters in one file

What it shows

PureEdgeSim is built from interchangeable models, and Simulation exposes one setter per model, but no other example registers all of them at once. This file is a tour of the whole wiring API, so the intent is different from every other custom example: it must prove that injecting a model that only observes changes nothing. Open extension_points.py and the eight set_custom_... calls at the end of the simulation() function.

Each registered class is a thin Tour... subclass that counts its own calls, prints the first one through record, and then delegates to the default it extends, so the scenario is untouched. The simulation is therefore launched twice, once with the stock models and once with the tour, and the two CSV rows must be identical. That equality is the point, a wiring operation that observes must be a no-op on the result, exactly like registering a Java default explicitly does nothing.

flowchart LR
    s_mgr["set_custom_simulation_manager"] --> t_mgr["TourSimulationManager"] --> d_mgr["SimulationManager"]
    s_net["set_custom_network_model"] --> t_net["TourNetworkModel"] --> d_net["DefaultNetworkModel"]
    s_gen["set_custom_computing_nodes_generator"] --> t_gen["TourNodesGenerator"] --> d_gen["ComputingNodesGenerator"]
    s_top["set_custom_topology_creator"] --> t_top["TourTopologyCreator"] --> d_top["DefaultTopologyCreator"]
    s_task["set_custom_task_generator"] --> t_task["TourTaskGenerator"] --> d_task["DefaultTaskGenerator"]
    s_mob["set_custom_mobility_model"] --> t_mob["TourMobilityModel"] --> d_mob["DefaultMobilityModel"]
    s_orch["set_custom_orchestrator"] --> t_orch["TourOrchestrator"] --> d_orch["DefaultOrchestrator"]
    s_node["set_custom_computing_node"] --> t_node["TourComputingNode"] --> d_node["ComputingNode"]
    classDef default fill:#cfe8ff,stroke:#333;
    class d_mgr,d_net,d_gen,d_top,d_task,d_mob,d_orch,d_node default;

The scenario

The settings folder ships the seven architecture by two algorithm sweep of the Java defaults (CLOUD_ONLY through ALL, TRADE_OFF and ROUND_ROBIN, 100 to 300 devices on a 200 by 200 map with one edge data center and one cloud). The script shrinks that to a single short run so the tour stays readable: it overrides to 20 devices, 300 simulated seconds, the ALL architecture, the TRADE_OFF algorithm, random_seed=7, and no log file. It then launches that one scenario twice, once stock and once with the eight tour classes chained onto the Simulation.

Run it

From the repository root, after pip install -e .:

python examples/extension_points/extension_points.py

The same script runs without installing the package, as long as src is on the path:

PYTHONPATH=src python examples/extension_points/extension_points.py

An optional device count and output folder shrink the tour further.

What to look for

extension point                class                      first use
simulation manager           TourSimulationManager    scenario ALL/TRADE_OFF/20
computing nodes generator    TourNodesGenerator       1 cloud, 1 edge, 20 devices
topology creator             TourTopologyCreator      84 directed links
task generator               TourTaskGenerator        1035 tasks
mobility model               TourMobilityModel        starts at (85, 176)
computing node               TourComputingNode        cloud 0 received its first event
orchestrator                 TourOrchestrator         task 451 -> wifi device 15
network model                TourNetworkModel         first transfer: task to destination
simulation manager           TourSimulationManager         1 uses
network model                TourNetworkModel           2070 uses
computing nodes generator    TourNodesGenerator            1 uses
topology creator             TourTopologyCreator           1 uses
task generator               TourTaskGenerator             1 uses
mobility model               TourMobilityModel            20 uses
orchestrator                 TourOrchestrator            291 uses
computing node               TourComputingNode          8053 uses

stock models                 tasks= 1035 ok=  751 avg_delay=0.2025   energy=11.5627  cpu=0.6342%
eight custom models          tasks= 1035 ok=  751 avg_delay=0.2025   energy=11.5627  cpu=0.6342%

same results with and without the custom classes: True

Files

Where to go next