Re: [ns] Goodput calculation for TCP connections...

[Huan Pham <huan@cs.jcu.edu.au>]

Yes, no of packets sent between node 0 and 6 is not a goodput. However, I
think we can not use the largest sequence number of received packet to
calculate the goodput either.

The reason is that, some packets get lost when being transmitted or droped at
queue while the largest sequence number is increased each time, when there's
new packet to send, no matter if the packet is transmitted successfully or
not.

For TCP traffic, if a packet get lost and is retransmitted, it uses a new
sequence number from the previous value (IS THAT CORRECT?)

If you use largest sequence number to calculate the goodput, you will get a
value that higher than the actuall goodput.

My solution is to use the total bytes received successfully to get the
goodput.


Tan Koan-Sin wrote:

> no of packets sent between node 0 and 6 is throughput no goodput.
> if the simulation time is long enough, the largest sequence number
> of received packet could be used to calculated the goodput.
>
> On Mon, May 28, 2001 at 09:28:15AM -0700, Debojyoti Dutta wrote:
> > look at the ns-user archive, i think there were a couple of discussions
> > ... easy way
> >
> > sample.awk
> > ----
> > BEGIN         {c=0}
> > $1=="r" && $3=="0" && $4=="6" $$ $5="tcp" {c++}
> > END   {print c;}
> > ----
> >
> > then awk -f sample.awk out.tr will give you no of packets sent between
> > node 0 and 6
> >
> > debo
> >