Lab 7: TCP and TCP-friendly Traffic

Instead of making TCP more aggressive, one can alternatively make UDP less aggressive, or behave in a TCP-friendly manner. In this exercise, you are asked to replace the multicast UDP connection with TFRC, one of the TCP-friendly control mechanisms implemented in ns-2, and see for yourselves whether TFRC is effective in the Lab 4 scenario in cutting the shares fairly between the TCP and non-TCP traffic.

You are also asked to take a closer look at the results and to make sure the conclusion drawn is reasonable.

7.a.Co-existing TCP and TCP-friendly Traffic

Take your lab4.tcl
Replace the multicast UDP connection with a TCP-friendly connection from node 7 to node 0.

ns-2 API to create a TCP-friendly connection (TFRC)

      set tfrc [new Agent/TFRC]
      set tfrcsink [new Agent/TFRCSink]
      $tfrc set class_ 2
      $tfrcsink set class_ 2
      $ns attach-agent $n(7) $tfrc
      $ns attach-agent $n(0) $tfrcsink
      $ns connect $tfrc $tfrcsink
      $ns at 0.2 "$tfrc start"

Insert above section of code into your lab4.tcl
Re-run the previous simulation and see if it actually improve TCP's bandwidth share.
Keep in mind that TCP-friendly traffic is now indicated as 'tcpFriend' in the trace, so change your awk script accordingly.
Is TFRC TCP friendly?

7.b.Removing Bias hidden in The Results

It appears that TFRC is using a very small share.
Check the ratio of TCP to non-TCP traffic in isolation.
- Comment out "$ftp start" and see how much non-TCP traffic there is supposed to be.
- Similarly, comment out "$tfrc start" (uncomment "$ftp start") and see how much TCP traffic there is supposed to be.
Look from the islolated case back to the mixed traffic case. Does the ratio of TCP (or non-TCP) traffic remain similar?
Is TFRC TCP friendly?

References:

The main VINT/ns page
The ns Manual page

Last modified: Thu Apr 26 20:08:57 MET DST 2001