As quoted from Curtis Villamizar:
> Eric Travis writes:
> > Everyone advertising arbitrarily
> > large windows are not a particularly good idea - at least right now
>
> btw- Arbitrarily large windows are just fine if sufficient buffering
> *and* RED gateways are used. One or the other is insufficient. I
> suppose this is stated without proof. Can I get away with "Proof is
> an exercise for the reader"?)
In the limited tests that I ran (single TCP stream) I found that as long
as the window is at least as large as the BDP, you get the maximum
TCP throughput. Even a window of 8*BDP shows no performance decrease.
So bigger is no problem.
[Note: I'm pretty sure the routers on both end don't do RED.]
Regards,
--
-- Gertjan van Oosten, [email protected], West Consulting B.V.
--
KAS-Netbank: <URL:http://www.kasnetbank.com/>, winnaar van de
!!! COMPUTABLE IT AWARD 1997: <URL:http://www.computable.nl/cptita1.htm> !!!
en de Computable Innovatieprijs 1997
This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:33 EST