Test steps

      1. Config IxNetwork with switches (first with 100 (ring) or 10 (full mesh) ) with ring or full mesh topology and two hosts in the adjacent switches (ring) or any other two switches (full mesh).

      2. Start the controller with the features install.

      3. Start the OF protocol of the switch.

      4. Wait until the channel is established and Echo message interaction started.

      5. Enable the traffic between this two hosts, wait until the hosts can ping through each other.

      6. Start the capture of IxNetwork.

      7. Disconnect the link of this two switches, and wait until the traffic recovery

      8. Stop the capture, analyse the messages captured. Write down the time of the port status message T1 and the time of the last flow mod message T2. Caculate the time of Datapath recovery time T=T2-T1

      9. Clean the configuration of controllers and IxNetwork.

      10. Repeat test step 1-9 for three times.

      11. Restart the test with another number of switches.

Test Results

• Datapath Failure Recovery Time (ring)

  Switches	First (ms)	Second (ms)	Third (ms)	Avg (ms)
  100	1381	951	857	1063
  200	2365	2128	1844	2112
  300	5556	5103	4531	5063
  400	10527	9063	12007	10532
  500	18281	22523	18142	19649
  600	35883	45297	35568	38916
  700	95209	76759	59439	77136

• Datapath Failure Recovery Time (full mesh)

  Switches	First (ms)	Second (ms）	Third (ms)	Avg (ms)
  10	702	704	857	739
  20	304	443	336	361
  30	488	334	341	388
  40	375	387	700	487
  50	311	691	706	569
  60	971	587	522	660
  70	512	483	721	572
  80						524	794	776	698

• Histogram of Datapath Failure Recovery Time
  Image Added
  Image Added

As the result shown, in ring topology, datapath failure recovery time increase significantly with the number of switches. But in full mesh topology, the time is no diffenence with the number of switches.