From OpenFlow Wiki
Status
12/10:
- Phase 0: success
- Phase 1.0 (running NOX): success
- Phase 1.2: success -- routing does exact match, including ARP L3 fields. (ARPs didn't get through until KK fixed the code to set L3 fields on ARP).
Components and Dependencies
- OpenFlow reference implementation
- NOX:
- depends on openflow ref
- need wireline compatible version
- FlowVisor:
- depends on openflow ref
- ETA Friday (?)
- Netfpga:
OpenFlow v1.0 Test Network
Move to OpenFlow v1.0 test network description page.
Phase 0 : Single switch
- Success
- Single soekris box in 3.2 closet
- controller from openflow reference implementation
- move Brandon (others?) over to it
- Start Monday 12/8 (??)
Phase 1 : Test Nox + Most features
- Add 2nd soekris box in 3.3 closet, physical wire that connects to 3.2
- Replace controller with NOX, initially running just routing (implicitly discovery, etc.)
- Move Masa over
- test slicing in an offline testbed (T-labs los altos)
- Phases 1.x might be able to happen in parallel
Phase 1.0 : Run NOX
Phase 1.1 : Test flow cookies
- modify NOX routing module to set/test flow cookie Done
Phase 1.2 : Test arp l3 matching
-
setup arp proxy and NOX module that forwards all arp for a certain subnet to it
- NOX uses exact match so IP in ARP must work for routing to work Done
Phase 1.3 : Test specify port in port stats
- test via dpctl Done
-
consider modifying envi to expose this
Phase 1.4 : Test match on ToS bits
- test with iperf (to set ToS for different packets) and dpctl (to set flows)
- Need to add ability to specify IP ToS bits to dpctl (Glen: end-of-week)
- Install static flow table entry that sends non-ToS packets to a particular port
- Only packets with ToS set will be sent to controller and thus routed
Phase 2 : Deploy slicing
- integrate FlowVisor
- add a slice for iperf to test slicing
- add soekris switch in flash lab
- move (some?) APs over
- move all rest of lab over to 1.0 switches
Phase 3 : Deploy on hardware
- blocking on openflow support for HP, NEC
- deploy 1.0 on hardware