using dummynet to simulate modem, dsl, etc
Luigi Rizzo
rizzo at icir.org
Fri Jul 25 10:38:39 PDT 2003
at the very least you are passing packets through each pipe
twice (once in ip_output(), once in ipo_input()),
which coupled with the delays causes a lot of interference in
the queues.
I would expect you to get around 3.5KBytes/s in the steady state,
but with a packet transmission time of 200ms and the 600ms RTT of
your setting, the first few rounds might achieve somewhere between
that and twice the bandwidth because of less interference.
This is exactly what you are seeing.
Try this:
ipfw -q add 1 pipe 1 ip from any to 127.0.0.100 in
ipfw -q add 2 pipe 2 ip from 127.0.0.100 to any in
ipfw -q pipe 1 config bw 33kbit/s queue 0 delay 40ms
ipfw -q pipe 2 config bw 56kbit/s queue 0 delay 112ms
to have a more realistic setting (300ms RTT is still a bit on the
high side i would say.)
cheers
luigi
On Fri, Jul 25, 2003 at 11:09:11PM +1000, Mark Hannon wrote:
> Hello,
>
> I am trying to use dummynet to simulate different network links and
> demonstrate
> application level performance for a variety of scenarios. The first
> test is three
> ftp file transfers over a simulated 56k modem as per below:
>
> iifconfig lo0 mtu 1500
> fconfig lo0 alias 127.0.0.100
> ipfw -q add 1 pipe 1 ip from any to 127.0.0.100
> ipfw -q add 2 pipe 2 ip from 127.0.0.100 to any
> ipfw -q pipe 1 config bw 33kbit/s queue 0 delay 40ms
> ipfw -q pipe 2 config bw 56kbit/s queue 0 delay 112ms
>
>
> The results seem strange ... larger file sizes seem to have a slower
> transfer rate,
> where I would have thought that steady-state transfers would have been
> reached
> for the larger files. Are these results logical? Have I done something
> silly with the
> pipe configs?
>
> Receiving test_large (1937201 bytes): 100% (ETA 00:00)
> 1937201 bytes transferred in 551.1 seconds (3.43 kBps)
> Receiving test_medium (248476 bytes): 100% (ETA 00:00)
> 248476 bytes transferred in 66.7 seconds (3.64 kBps)
> Receiving test_small (40960 bytes): 100%
> 40960 bytes transferred in 7.3 seconds (5.52 kBps)
>
>
> Regards/mark
>
>
>
>
>
>
> _______________________________________________
> freebsd-ipfw at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
> To unsubscribe, send any mail to "freebsd-ipfw-unsubscribe at freebsd.org"
More information about the freebsd-ipfw
mailing list