Re: [ns] tcp bw

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

On Thu, 16 Nov 2000, betty Royer wrote:

> Why TCP does'nt use all the 

path capacity

> bw when the propagation delay is
> long. For exmple : I used only one source of TCP over the
> link 100ms 2Mb and the bw of TCP was only 800Kb. it's not
> the same when i use 10ms, there tcp take all the bw.

assuming no errors/losses on the path and that you're running
sikmulations for long enough for TCP to fill capacity, look at the
window size:

ns-default.tcl:Agent/TCP set window_ 20

The sender has to store a copy of what's in flight until it knows it's
been received successfully; if there's a longer flight and more to
store in the same space, this limits the amount of new data that can
be accepted from the application and sent out. ns sets the size in
terms of segments.

(the receiver has a similar buffer for reassembling segments, but that
really only comes into play when losses/reordering/SACK crop up.)

Read Stevens vol. 1. Long-delay paths are done to death by
geostationary satellite enthusiasts;
C. Partridge and T. Shepard, "TCP Performance over Satellite
Links," IEEE Network, vol. 11, no. 5, pp. 44--49, Sept. 1997.
is a good overview.

W. Stevens, TCP/IP Illustrated, Volume 1, Addison Wesley, 1994.

L.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>