- The idea is to route ICMP traffic in the emulated Stanford-backbone network.
- Follow the configuration instructions to generated the necessary OpenFlow rules (translated from the Stanford-backbone network).
- Follow the controller instructions to setup the controller which would install the previously-generated OpenFlow rules into Mininet.
- Follow the topology instructions to setup the Stanford-backbone topology in Mininet. Start controller before Mininet. Otherwise exisiting dummy flow entries might be cleared.
- From mininet net output(
mininet> net) we noticed the connection:h197(h197-eth0) --- (s13-eth9)s13(s13-eth8) --- (h196-eth0)h196. So h197 is where we want to initiate the ICMP traffic. - Check what routes are available at s13 by
sudo ovs-ofctl dump-flows s13. Observe this:cookie=0x0, duration=3189.633s, table=0, n_packets=0, n_bytes=0, priority=59929,ip,nw_dst=128.12.1.248/29 actions=output:5,output:8,output:2. So we can try to ping128.12.1.249. And if things work, the ICMP traffic should be forwarded to boths13-eth2ands13-eth8but not to any other ports. - Prepare h197 to initiate ICMP traffic.
- View the route table by
h197 route -n. Make sure the default route is present. - Setup static ARP by
h197 arp -s 128.12.1.249 01:00:00:00:00:23. Without this, the PING client could not figure out what MAC address to use, i.e. some ARP protocol is required.
- View the route table by
- Start ICMP traffic:
h197 ping 128.12.1.249. - Check if the ICMP traffic is forwarded properly.
- Check if s13 receive the traffic:
sudo tcpdump icmp -i s13-eth9 -nS. You should see ICMP traffic. - Check if s13 forward the traffic correctly:
sudo tcpdump icmp -i s13-eth2 -nS&sudo tcpdump icmp -i s13-eth8 -nS. You should see ICMP traffic.sudo tcpdump icmp -i s13-eth3 -nS. You should not see ICMP traffic (and for all other ports except 2 & 8). - Ping should fail. Because we did not configure host IP.
- Check if s13 receive the traffic:
- Prepare h197 to initiate ICMP traffic to Google (74.125.226.78).
- Probe the path
- Use the topology information to figure out peering ports.
- See what flow table entries are being matched by invert matching:
sudo ovs-ofctl dump-flows s13 | grep -v n_packets=0
- The path of Google traffic:
**
h197-eth0 > s13-eth9 > s13-eth1 > s1005-eth4 > s1005-eth1 > s2-eth13 > s2-eth4 > s1009-eth1 > s1009-eth2(s1009-eth3) > s3-eth1(s12-eth9) > s3-eth7(s12-eth1) > h67-eth0(s1007-eth3) > (s1007-eth1) > (s1-eth1) > (s1-eth4) > (h18-eth0)8* It seems that the traffic is forwarded to two destinations. One is a border router (s1 is the bbra_rtr). Another is an operational zone router (s3 is boza_rtr).
- Prepares h197 to initiate ICMP traffic to
stanford.cs.edu(171.64.64.64). - Probe the path:
h197-eth0 > s13-eth9 > s13-eth6 > s1006-eth3 > s1006-eth1 > s1-eth32 > s1-eth22 > h33-eth0. - Inject fault:
- Add a dropping rule by
sudo ovs-ofctl add-flow s1 dl_type=0x0800,nw_dst=171.64.64.64/32,priority=65535,actions= - The previous traffic will be dropped. Run
sudo tcpdump -i s1-eth32 -nSyou should still see traffic. However, runsudo tcpdump -i s1-eth22 -nS, and you should not see traffic. Bascically, traffic is dropped at s1. - How to check if packets are received: If we dumped the traffic at at s1-eth22 (pcap), we can see that traffic stops at the 58173th packet. We could filter the time interval by
editcap -A "2014-8-11 13:51:28" -B "2014-8-11 13:52:00" s1-eth22.pcap time.pcap. Then we can see if packets of interest did arrive:time tshark -r time.pcap -R "(icmp and ip.src!=10.0.0.17)" - If we dumped roughly 10 minutes of traffic (roughly 50 GB of headers):
- The experiment happens on August 13, starts at 15:40, drops at 16:19, ends at 16:21.
- No packet at s2017-eth1 (0.51 + 0.11 = 0.62 second).
- No packet at s2017-eth2 (0.51 + 0.11 = 0.62 second).
- No packet at s1-eth22 (0.49 + 0.11 = 0.60 second).
- Packet at s1-eth32 (0.82 + 0.22 = 1.04 second).
- Packet at other interfaces at s1 (estimated 6x more traffic, so 1.04 * 6 = 6.24 second).
- Total pcap query latency is roughly 10 seconds. We touched roughly 15% of all the captured pcap traces.
- Add a dropping rule by