[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: asymetric link
I have tried your example. I can see the same delay for both pkt and ack,
too. It seems that NS is not doing properly the network description
for NAM:
If I set a new node in the middle
28k 28k
------- n2 --------
/ \
n0 +---------------------+ n1
10M
Now it works; I do see the difference between one path and the other.
Taking a look to the nam output files for the first and second examples,
we see that in the second case we have three lines like this in out.nam:
l -t * -s 0 -d 1 -S UP -r 10000000 -D 0.59999999999999998 -c black -o
l -t * -s 1 -d 2 -S UP -r 28800 -D 0.01 -c black -o
l -t * -s 2 -d 0 -S UP -r 28800 -D 0.01 -c black -o
I don't know many things about NAM but this seems to be the specification
of the links, with their bandwidth and delay.
In the first case (the example you posted), you can't find in nam.out a
line like:
l -t * -s 1 -d 0 -S UP -r 28800 -D 0.01 -c black -o
(In fact you will not find the keyword "28800" in all the file).
If you add it manually, you get a "duplicate edge" error message (I don't
know if I'm doing it properly). I wonder if NAM can do such a kind of
graphs...
I append both the files. out.nam.orig and out2.nam. Maybe a NAM guru can
tell us what happens.
/FCD
out.nam.orig:----------------------
V -t * -v 1.0a5 -a 0
A -t * -n 1 -p 0 -o 255 -c 15 -a 1
A -t * -h 1 -m 127 -s 8
c -t * -i 0 -n Blue
c -t * -i 1 -n Red
n -t * -a 0 -s 0 -S UP -v circle -c black
n -t * -a 1 -s 1 -S UP -v circle -c black
l -t * -s 0 -d 1 -S UP -r 10000000 -D 0.59999999999999998 -c black -o
+ -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 ------- null}
- -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 ------- null}
h -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 -1 ------- null}
r -t 0.600112 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 -------
null}
+ -t 0.600112 -s 1 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
- -t 0.600112 -s 1 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
h -t 0.600112 -s 1 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
r -t 0.621223 -s 1 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
----------------out.nam.3nodes:--------------------
V -t * -v 1.0a5 -a 0
A -t * -n 1 -p 0 -o 255 -c 15 -a 1
A -t * -h 1 -m 127 -s 8
c -t * -i 0 -n Blue
c -t * -i 1 -n Red
n -t * -a 0 -s 0 -S UP -v circle -c black
n -t * -a 1 -s 1 -S UP -v circle -c black
n -t * -a 2 -s 2 -S UP -v circle -c black
l -t * -s 0 -d 1 -S UP -r 10000000 -D 0.59999999999999998 -c black -o
l -t * -s 1 -d 2 -S UP -r 28800 -D 0.01 -c black -o
l -t * -s 2 -d 0 -S UP -r 28800 -D 0.01 -c black -o
+ -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 ------- null}
- -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 ------- null}
h -t 0 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 -1 ------- null}
r -t 0.600112 -s 0 -d 1 -p tcp -e 140 -c 0 -i 0 -a 0 -x {0.0 1.0 1 -------
null}
+ -t 0.600112 -s 1 -d 2 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
- -t 0.600112 -s 1 -d 2 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
h -t 0.600112 -s 1 -d 2 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
r -t 0.621223 -s 1 -d 2 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
+ -t 0.621223 -s 2 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
- -t 0.621223 -s 2 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
h -t 0.621223 -s 2 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
r -t 0.642334 -s 2 -d 0 -p ack -e 40 -c 0 -i 1 -a 0 -x {1.0 0.0 -1 -------
null}
>
> When I run the nam, I notice that the speed of the first
> emitted pkt is equal to the speed of the corresponding ack,
> which should NOT be the case because the link is asymetric.
> Is this a problem related to the link, or to TCP (do I have
> to use another TCP ?) ?
>
> Thank you very much, and sorry for disturbing
>
> Karim
> ______________________________________________________
> Bo�te aux lettres - Caramail - http://www.caramail.com
>
>