[8698] in Commercialization & Privatization of the Internet
Re: One-way Internet?..no, it's "Asymetric"
daemon@ATHENA.MIT.EDU (Dick St.Peters)
Tue Nov 30 20:13:22 1993
Date: Tue, 30 Nov 93 20:06:44 EST
From: stpeters@spare-parts.crd.ge.com (Dick St.Peters)
To: com-priv@uu.psi.com, emv@garnet.msen.com
Reply-To: <stpeters@dawn.crd.ge.com>
> From: emv@garnet.msen.com (Edward Vielmetti)
> You need enough bandwidth to be able to ack packets at maximum rates.
> We went through this calculation once before - I think Craig Partridge
> came up with the numbers - such that if you really have 10mb/sec
> ethernet going in one direction you only get to use a small part of
> it when the acks come back on a 14.4 modem. Using something like
> ISDN for the return channel makes a substantial difference in throughput.
Ed, Craig started that discussion, but he didn't take into account
header compression. The number we wound up with is that the incoming
bandwidth can be a little over 400 times the outgoing bandwidth.
Hybrid's claim is 413 times. With 14.4 Kbaud going out, that gives a
hair under 6 Mbps coming in, nominally. ISDN going out runs the
in-bandwidth up to the 10 Mbps limit, a "substantial difference in
throughput" only if you're talking to someplace that can pump you data
at these rates.
--
Dick St.Peters
GE Corporate R&D, Schenectady, NY stpeters@dawn.crd.ge.com
