SDN-based Intelligent Network Controller

Built a software‑defined networking (SDN) controller to explore control‑plane programmability and performance trade‑offs.

  • Learning Switch: Proactive MAC learning + flow pre-installation achieving 18.5 Gbit/s (≈670× reactive flood baseline).
  • Topology Awareness: LLDP-based link discovery feeding dynamic adjacency graph.
  • Routing: Dijkstra shortest paths with link weights derived from active delay measurements; installs per‑path flow rules.
  • Loop Avoidance: Implemented STP-like pruning to prevent broadcast storms in cyclic graphs.
  • Monitoring: Periodic counters + congestion heuristic to trigger path recomputation.

Results: Significant latency reduction and throughput gains under synthetic multi-host workloads; validated scalability on fat-tree emulations.

Tech: Python, Ryu, Mininet, OpenFlow.

View on GitHub →

Controller exposes a metrics endpoint exporting flow install latency and path recomputation intervals for evaluation.