Guido Reina wrote:
>
> Thanks for your explanations. The problem we have is slighty
> different. We have different lines with differents Qualitys. The
> problem is , if the line has a lot of noise (Poor Quality) for us is
> better to copy by small fragments instead big chunks. In fact the
> final idea is to calculate in real time the number of failed attemps
> and then if one of the destinations achieved a lot of them, then try
> to adjust the buffers to low size to avoid the noise in the line. This
> well be done dinamically.
>
> Thanks a lot for your pattient, but as you see this problem is very
> specific and we are working with this kind of problems during several
> years and i see if i do it manually (To reduce the buffer) it's works
> really fine.

Do you know which lines are of poor quality and which are not? If so,
then on the sender, add specific host/network routes for the
networks/hosts reached through those lines, specifying a smaller Path
MTU. TCP will then use a smaller segment size, but you can keep the
large TCP window.

You get the smaller TCP segments which will be less likely on average to
have a bit error in them, but can keep the window sized such that you
have a decent shot at getting a fast retransmit

Another way to go is at the router for each of those noisy lines, reduce
the MTU and let "normal" Path MTU discovery give you the small TCP
segment size.

rick jones

-- 
Wisdom Teeth are impacted, people are affected by the effects of events.
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to raj in cup.hp.com  but NOT BOTH...
-------------------------------------------------------
This sf.net email is sponsored by: Influence the future 
of Java(TM) technology. Join the Java Community 
Process(SM) (JCP(SM)) program now. 
http://ads.sourceforge.net/cgi-bin/redirect.pl?sunm0004en