EE555-Openflow
Topic: Openflow Based SDN Controller Design using Pox Library
Protocols Concerned: Arp, ICMP, Ethernet, IPv4
Commands to Run Scenario 1 ~ 4.5
- Please start Pox before Mininet to make sure controllers are discoverable
- Scenario 1: Flow Installation Tutorial
- Start Mininet: sudo mn --topo single,3 --mac --switch ovsk --controller remote
- Start Pox: ./pox/pox.py log.level --DEBUG misc.of_tutorial
- Scenario 2: Arp & ICMP Handling In Single-Host Subnets & Single Switch
- Start Mininet: sudo mn --custom topology2.py --topo mytopo --mac --controller remote
- Start Pox: ./pox/pox.py log.level --DEBUG misc.controller2 misc.full_payload
- Scenario 3: Multi-Host Subnets & Multiple Switch
- Start Mininet: sudo mn --custom topology3.py --topo mytopo --mac --controller remote
- Start Pox: ./pox/pox.py log.level --DEBUG misc.controller3 misc.full_payload
- Scenario 4: Looped Connection Between Switches
- Start Mininet: sudo mn --custom topology4.py --topo mytopo --mac --controller remotee
- Start Pox: ./pox/pox.py log.level --DEBUG misc.controller4 misc.full_payload
- Scenario 4.5: Firewalling TCP Packets
- Start Mininet: sudo mn --topo single,3 --mac --switch ovsk --controller remote
- Start Pox: ./pox/pox.py log.level --DEBUG misc.firewall misc.full_payload
Comments:
- Scenario 2 & 3 provides *_messy.py & *_neat.py of the controller.
- They both work, while ICMP & TCP/UDP situations in *_neat.py are jointly handled
- *_messy.py is served for convenience during development
- Code quality are improved moving from Scenario 2 -> 3
- Including better function APIs & pre-fetching informations
- Majority part of Scenario 4 is identical to that of Scenario 3's
- Only routing tables are changed
- Arp Request Forwarding is done in iterative manner
- Expected TCP bandwidth using "iperf hi hj": 40+ Gbits/s
- UDP tests are much slower for unknown reasons