Re: [ns] Study of TCP

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

On Tue, 7 Mar 2000, Ana Delgado wrote:

> I'm working with TCP at the moment and I have problems to get some 
> parameters in the simulation, do you know how can I get:
> 1. The bytes of the congestion window vs time.
> 2. The sequence number vs time.

More tracefile howto documentation is needed; I keep having to walk
people through the idea that they can write a procedure to record what
interests them, and then a lightbulb goes on...

proc record {} {
         global ns tcp0 trace_wnd trace_ack trace_seq trace_srtt
         
         set time 0.01
 
         #Get the variables of interest
         set curr_dupack [$tcp0 set dupacks_]
         set curr_cwnd [$tcp0 set cwnd_]
         set curr_maxseq [expr fmod([$tcp0 set maxseq_], 100)]
         set curr_srtt [$tcp0 set srtt_]
         #Get the current time
         set now [$ns now]
         #Write the variables of interest vs. time
         puts $trace_ack "$now $curr_dupack"
         puts $trace_wnd "$now $curr_cwnd"
         puts $trace_seq "$now $curr_maxseq"
         puts $trace_srtt "$now $curr_srtt"
 
         #Re-schedule the procedure
         $ns at [expr $now+$time] "record"
}

[..]

set trace_wnd [open ngsonewreno.wnd w]
set trace_ack [open ngsonewreno.dup w]
set trace_seq [open ngsonewreno.seq w]
set trace_srtt [open ngsonewreno.srtt w]

[..]

$ns at 4.0 "record"

[..]

$ns at 100.0 "end"

proc end {} {
    global ns trace_ack trace_wnd trace_seq trace_srtt
    close $trace_wnd
    close $trace_ack
    close $trace_seq
    close $trace_srtt
}

multiply curr_cwnd by the segment size to get it in bytes; in ns TCP
the latter is unchanging.


> 3. The losses segments vs time.

look for 'd' events in the trace-all tracefile.


> 4. The RTT used by every packet number.

Analyse trace-all tracefile to get packet/ack rtts.


> 5. The relative bandwidht vs time.

Analyse trace-all tracefile.

L.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>