[ns] TCP and packet drop: where are the retransmitted packets?
arne.lie at item.ntnu.no
Tue Jun 27 07:16:47 PDT 2006
Actually; I just discovered that the retransmitted packets ARE there in the
trace file. The TCP sequence number (column two from the right) is used as
it should, in addition an "A" flag is used to mark a retransmitted packet.
Something went wrong with my first search on this topic. Again, ns-2 is my
best friend, and thanks to Phil and Ethan for their time. - Arne
> -----Opprinnelig melding-----
> Fra: unmobile at gmail.com [mailto:unmobile at gmail.com] På vegne
> av Phil Miller
> Sendt: 27. juni 2006 14:30
> Til: Arne Lie
> Kopi: ns-users at isi.edu
> Emne: Re: [ns] TCP and packet drop: where are the
> retransmitted packets?
> >From what I've observed, the normal TCP agents don't
> actually reliable
> send each packet in the stream, they only simulate the flow
> and congestion control dynamics that TCP would exhibit, so
> that the amount of generated traffic is accurate. Give the
> FullTCP agents a try.
> On 6/27/06, Arne Lie <arne.lie at item.ntnu.no> wrote:
> > Hi all,
> > For some time I have only been interested in the congestion
> > of the different TCP congestion control algorithms, and the way it
> > balances flow throughput to available capacity, packet drop
> statistics, etc.
> > However, now I need to see how well TCP tackles
> retransmissions, e.g.
> > with SACK. When I inspect the event trace file, dropped
> packets are "never"
> > retransmitted, in that ns-2 seems to increment the sequence numbers
> > (both per flow and per simulation) all the time,
> > >>>>>>>making it impossible to know WHEN a retransmission
> is performed.
> > <<<<<<<<<
> > Is this a correct observation? If yes, are there any
> workarounds? If
> > incorrect, how *do* I find the retransmission event, and e.g. how
> > *many* packets are duplicated/retransmitted (SACK vs. non-SACK)?
> > best regards
> > Arne
More information about the Ns-users