Closing TCP connections

[Christopher Clark <cwc22@hermes.cam.ac.uk>]

Hello

Using the standard one way TCP implementation, I have the below code to
simulate many TCP connections being initiated over time from a set of
source hosts to a set of sinks. I am concerned that I may not be causing
connections to be closed properly, meaning that if I run a larger
simulation, the maximum number of connections from source to sink will be
reached - is this possible? 

If I use FullTCP, what would cause a FIN packet to be sent?

Any help would be most appreciated.

Christopher Clark



(20 source hosts named h1 to h20; twenty sinks name sk1 to sk20)

(initial test code:)
  set hnum 1
  while { [info exists node_(h$hnum)] } {
        set fid_(h$hnum) 1
        $self hostKick $hnum 1
        incr hnum
  }

...

Test/transatlantic instproc hostKick { hnum sknum } {
        $self instvar ns_ node_ rng_ fid_
        if [info exists node_(h$hnum)] {
                # this ensures the 'kicks' cycle through the target hosts
                set skdest $sknum
                incr sknum
                if ![info exists node_(sk$sknum)] {
                        set sknum 1
                }

                # we're taking the mean delay between kicks at .5 second
                # (exponentially distributed, mean rate of connections per
host 0.5s)

                set delay [expr [$rng_ exponential] / 2]
                # schedule the next kick for this host
                # this delay is also used below as the start time of
		# connection
                $ns_ at [expr [$ns_ now] + $delay] "$self hostKick $hnum
$sknum "
                
                set fid $fid_(h$hnum)
                set tcplocal [$ns_ create-connection TCP $node_(h$hnum)
TCPSink $node_(sk$skdest) $fid]
                incr fid_(h$hnum)

                set ftplocal [$tcplocal attach-source FTP]
                $ns_ at [expr [$ns_ now] + $delay] "$ftplocal produce
1000"

        }
}