Page History

Second set of experiments was set up on testbed 1975. There was cross traffic between H6 and H2 (ping with 1000 bytes packets with 0.2s interval between them) on both LSP with MPLS label 21 and LSP with MPLS label 22.

The packets were captured on H5, H4, H3 and H2 on eth2 links (ingress for echo requests).

In parallel, the d-cart system was used between H6 and H2 in order to verify the capturing results (one directed flow on the LSP whose MPLS Label is 22).

Packets have been captured on H2 and the path monitored with D-cart for approx. one minute in both experiments.

A tshark instance was also running on H4 (here are the raw traces H4-tshark.zip and the command is, e.g.:

sudo tshark -i eth1 -Tfields -E separator='|' -e mpls.label -e vlan.id -e eth.type -e frame.time_epoch -e ip.src -e ip.dst -e ip.id -e icmp.seq -e tcp.seq -e ip.checksum).

The first set of results is here (captures are in 4 different formats, sorted by the MPLS label):

Capture from h2 no netem.rar 

Capture from  h3 no netem.rar

Capture from  h4 no netem.rar

Capture from h5 no netem.rar  

D-cart results for this setup are here  d-cartH2wonetem      

The format is the following:

  code timestamp seq_num delay (ip.id)

  code := 1 (loss or unordered packets), 2 (#hop change - not used in gts), 4 (significant delay change), 8 (heartbeat - every 100 packets)

 if the code is even then you have a H(istory) with the last packet received and its own info...

The sending calibration is: 1 dcart ping (smallest possible packets but "full" of extra info) each 100ms

In the second setup, netem tool was used on the link H4-H3 (eth1 on H4) with the following parameters: delay 100ms 2ms distribution normal loss 10%

The set of results of the measurements with netem switched on is here:

Capture from h2 with netem.rar 

Capture from  h3 with netem.rar

Capture from   h4 with netem.rar

Capture from h5 with netem.rar