[ns] TSW2CM vs. Token bucket

[Sungwon Yi <syi@cse.psu.edu>]

Dear ns-users,

Hi, I found the strange result from the simulation in ns, and I used
following topology.

n(1 to 10 ) - e1 \                         / d(1-10)
                  e3  ----  core  ---- e4 
n(11 to 21) - e2 /                         \ d(11-20)

1. I used TSW2CM in e1 and e2, and started 10 TCP flows.
   The cir used is 64000( 8Kbyte/sec ), packet size is 1k.
   The simulation lasts for 100sec, so that I expected 8000 of In-Profile
   packets transfered, but I got only around 7500 of In-Profile packets.
   
2. I changed TSW2CM to tokenbucket with same cir and cbs = 40000.
   At this time, the In-profile packets were around 8000.
   
Does this difference comes from the diffent policy of these two meters for
bursty traffic(TCP)?

I would greatly appreciate for any help.

Thanks
Sungwon.

======================================
Yi, Sungwon
Have a nice day!
======================================