My estimate was that if the window is 8192 bytes..commonly seen,
and assuming a 1 way time of 120 msec...worst case
each TCP flow has to be given a bandwidth of:
8192/(8*120 * 10^-3) = 8533.33 bps = 8.5 kbps end to end
This ofcourse does not include the fact that we put multiple headers on the
packet.
try using a lynx browser and browse on a 12kbps line and see if i am correct
;-)...
Does anyone have other numbers?
----- Original Message -----
From: "Alok" <alokdube@xxxxxxxxxx>
To: "Ethereal user support" <ethereal-users@xxxxxxxxxxxx>
Sent: Wednesday, September 21, 2005 6:35 PM
Subject: Re: [Ethereal-users] Throughput for a TCP connection?
> :-) perhaps you may like to ask this question to people at isi.edu
> we have to live with what is made...
> ----- Original Message -----
> From: "JS" <fedevaps@xxxxxxxx>
> To: <ethereal-users@xxxxxxxxxxxx>
> Sent: Wednesday, September 21, 2005 1:20 PM
> Subject: [Ethereal-users] Throughput for a TCP connection?
>
>
> > I have just oploaded a small text file to a server
> > where I used ethereal to capture the traffic. I have
> > read that I can calculate the average throughput for a
> > TCP connection like this:
> >
> > (0.75 * W)/RTT
> >
> > Where W is the increased window size by 1 MSS each RTT
> > until a loss occurs. But what if no loss occurs?
> >
> > I have quite many TCP packages with a lot of different
> > window sizes. If I need to do the calculation for the
> > whole TCP connection it will take most of the day
> > since RTT and window size varies all the time. What is
> > the good thing about this formular when I have window
> > sizes and RTT that varies?
> >
> >
> > _______________________________________________
> > Ethereal-users mailing list
> > Ethereal-users@xxxxxxxxxxxx
> > http://www.ethereal.com/mailman/listinfo/ethereal-users
> >
>
>
> _______________________________________________
> Ethereal-users mailing list
> Ethereal-users@xxxxxxxxxxxx
> http://www.ethereal.com/mailman/listinfo/ethereal-users
>