Re: About "increasing TCP's initial window"

From: Hans Kruse ([email protected])
Date: Tue Jan 13 1998 - 08:52:46 EST


At 2:53 -0500 1/13/98, kim yong sin wrote:
>By draft, Initial window size would be that given :
> min(4*MSS, max (2*MSS, 4380 bytes))
>
>Why don't consider 'RTT' ?

Kim, RTT is of course the reason why slow-start is more detrimental to
performance on high-speed geo satellite links. The problem I have is, how
would you know the RTT at connection start? Are you thinking about
persistent TCP connection information ("we have talked to this host before
and it is 500ms away")? Hmm, I suppose the three-way handshake could
provide a measurement, but is it reliable enough?

Hans Kruse, Associate Professor and Director
McClure School of Communication Systems Management, Ohio University
9 S. College Street
Athens, OH 45701
614-593-4891 voice, 614-593-4889 fax, [email protected]



This archive was generated by hypermail 2b29 : Mon Feb 14 2000 - 16:14:35 EST