A centralized SDN controller with distributed switches communicating over UDP. The controller computes shortest-path routing tables using Dijkstra's algorithm and pushes them to switches. Switches exchange KEEP_ALIVE heartbeats with neighbors and report topology changes to the controller, which recomputes routes only when the network topology changes. All messages use a struct-based binary protocol.
Launch an entire network (macOS only, opens separate Terminal windows):
python3 run_network.py <port> <config_file>
Example:
python3 run_network.py 9000 Config/graph_3.txt
Start the controller and each switch individually. The controller must be started first.
python3 controller.py <port> <config_file>
python3 switch.py <switch_id> <controller_host> <controller_port>
Example: run a 3-switch network:
python3 controller.py 9000 Config/graph_3.txt
python3 switch.py 0 localhost 9000
python3 switch.py 1 localhost 9000
python3 switch.py 2 localhost 9000
Simulate a link failure with the -f flag. The switch will not send or process KEEP_ALIVE messages to/from the specified neighbor:
python3 switch.py <switch_id> <controller_host> <controller_port> -f <neighbor_id>
Example:
python3 switch.py 0 localhost 9000 -f 1
This simulates a failed link between switch 0 and switch 1. After TIMEOUT (6s), both sides detect the dead link and the controller recomputes routes. To simulate a full switch failure, kill the switch process (Ctrl+C). Restarting a switch with the same ID will re-register it with the controller and rejoin the network.
Run perf.py alongside the network to log message rates, estimated bandwidth, and propagation delays to Performance.log:
python3 perf.py <config_file> [--interval SECONDS]
The default reporting interval is 10 seconds. With the automated launcher, pass -p:
python3 run_network.py 9000 Config/graph_3.txt -p
Each switch sends a REGISTER_REQUEST to the controller on startup. Once all switches have registered, the controller responds with neighbor information and computes initial routing tables using Dijkstra's algorithm.
Every 2 seconds, each switch sends KEEP_ALIVE messages to its alive neighbors and a TOPOLOGY_UPDATE to the controller. If no KEEP_ALIVE is received from a neighbor for 6 seconds, the link is marked dead and the controller is notified. The controller also monitors for full switch death if no TOPOLOGY_UPDATE arrives within 6 seconds.
The controller caches routing tables and only recomputes when the topology hash changes. Updated routes are pushed to all alive switches.
Each switch writes to switch<id>.log and the controller writes to Controller.log. Logged events include register requests and responses, neighbor and switch dead/alive transitions, link failures, and routing table updates.