Custom computing node: a node that handles an event of its own

What it shows

Other examples replace a policy or a model, but this one reaches the node itself, the object that every layer is built from. The class registered with set_custom_computing_node constructs the cloud data center, the edge data center, and every edge device alike, so a single subclass adds behaviour to the whole fleet. Open ReportingNode.

ReportingNode extends ComputingNode and adds exactly one thing: in on_simulation_start it calls schedule_now(self, DO_SOMETHING) to send an event to itself. In process_event it intercepts that event, counts it, prints, and returns, and forwards everything else to super().process_event. The private tag DO_SOMETHING = 12000 is the crux: the framework’s own tags are EventType members numbered 1 to 18 (EXECUTE_TASK is 3, UPDATE_STATUS is 15, DOWNLOAD_CONTAINER is 18), so any other integer is free, and it must not collide, because reusing a tag inside 1 to 18 would silently swallow a real task or status event. The type setter is there so all three layers still report their own NodeType and derive is_peripheral, which the mobile devices need to find their data centers.

flowchart TD
    start["on_simulation_start"] --> sched["schedule_now DO_SOMETHING 12000"]
    sched --> recv["process_event"]
    recv -->|"tag 12000"| own["count, print, return"]
    recv -->|"tag 1 to 18"| sup["super().process_event"]
    sup --> builtin["task and status events"]
    classDef own fill:#cfe8ff,stroke:#333;
    class sched,recv,own own;

The scenario

The settings folder carries the seven architecture by two algorithm sweep of the Java defaults, and the script overrides it to one short run: 20 devices on a 200 by 200 map, one cloud data center, one edge data center (dc1 at 100,100), the ALL architecture, the TRADE_OFF algorithm, random_seed=7, and 300 simulated seconds. That is 1 cloud node, 1 edge node, and 20 device nodes, 22 nodes in total, each of which receives the custom event once.

Run it

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

python examples/custom_computing_node/custom_computing_node.py

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

PYTHONPATH=src python examples/custom_computing_node/custom_computing_node.py

An optional device count and output folder shrink the run.

What to look for

Event received by cloud 0, you can do any action here
Event received by dc1, you can do any action here
Event received by wifi device 3, you can do any action here
Event received by wifi device 4, you can do any action here
Event received by wifi device 5, you can do any action here
Event received by wifi device 6, you can do any action here
Event received by wifi device 7, you can do any action here
Event received by wifi device 8, you can do any action here
Event received by wifi device 9, you can do any action here
Event received by wifi device 10, you can do any action here
Event received by wifi device 11, you can do any action here
Event received by wifi device 12, you can do any action here
Event received by wifi device 13, you can do any action here
Event received by wifi device 14, you can do any action here
Event received by wifi device 15, you can do any action here
Event received by wifi device 16, you can do any action here
Event received by wifi device 17, you can do any action here
Event received by wifi device 18, you can do any action here
Event received by wifi device 19, you can do any action here
Event received by wifi device 20, you can do any action here
Event received by wifi device 21, you can do any action here
Event received by wifi device 22, you can do any action here

22 nodes received the custom event: cloud=1, edge_datacenter=1, edge_device=20
custom_computing_node      tasks= 1035 ok=  751 avg_delay=0.2025   energy=11.5627  cpu=0.6342%

Files

Where to go next