[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Retransmissions
On Wed, 21 Jul 1999, Kartik Sudeep wrote several times:
> Date: Wed, 21 Jul 1999 13:15:49 PDT
> From: Kartik Sudeep <[email protected]>
> To: [email protected]
> Subject: Retransmissions
>
> Lloyd,
>
> Is there a way that I can find out the number of retransmitted packets in
> a two-way TCP from a trace file ?
I'm not the person to ask about the intricacies of ns TCP, much less
FullTcp. The rest of this list is better qualified.
Anyway, I'd go and modify Trace::format() to display tcph->reason().
(see nsDoc 18.4 and 24.4)
After that it's just a matter of counting the number of lines of
enqueued (+) packets at the source node going from the packet source
address (source node.blah) of the TCP sender, and subtracting the
number of lines where reason isn't zero - 1 is timeouts, 2 is dupacks,
3 indicates a rate-based-pacing hack to vegas which shouldn't appear.
And ~ns/tcl/ex/tcp-fs is filled with handy awk scripts to hack about
to do this, although, frankly, in this case those are overkill for
what can be done with a single simple grep -c -E pattern match that I
can't quite think of right now.
I suppose you could look for duplicate seqno_ and build lists of first
occurrences, but reason_ looks far far easier to parse when
post-processing the tracefile to me. Surprised it's not outputted
already, which probably means that I'm missing something subtle and
there's probably a much much simpler way to do this.
Like I said, I'm not the person to ask about ns TCP.
L.
spends lots of time annotating nsDoc tracefile excerpts with the field
positions as headings when he's not grumbling about how thinking in
perl completely messed up his regexps.
> -Kartik
>
>
> ______________________________________________________
> Get Your Private, Free Email at http://www.hotmail.com
>
<[email protected]>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>