[11440] in Commercialization & Privatization of the Internet
Re: Use market forces to deal with routing questions & charging
daemon@ATHENA.MIT.EDU (Dick St.Peters)
Sat Apr 2 22:51:09 1994
Date: Sat, 2 Apr 94 18:03:33 EST
From: stpeters@bird.crd.ge.com (Dick St.Peters)
To: jlw@cs.columbia.edu
Cc: com-priv@psi.com
Reply-To: <stpeters@dawn.crd.ge.com>
>From: James Waldrop <jlw@cs.columbia.edu>
>Sean Donelan wrote:
>>Unfortunately the costs to providers of the pipe don't work that way.
>>Generally they charge "individual" users a less assuming an some average
>>level of use, and size their networks on that basis. Resellers have a
>>different usage profile (usually greater), and therefore are being charged
>>different rates based on their average use.
>
>If I'm using 98% of my 56K link to Sprint, without reselling, are they
>losing money on me? Somehow I don't think so. And if they are, that's
>their business problem, not mine. At any rate, I don't see how CIX
>enters this picture at all.
*One* fully-utilized 56K link probably doesn't make a big difference to
Sprint, but Sean's point is valid. Full utilization of a lot of them
is a different matter. A T1 can mux 24 56's, but if an ISP can assume
most of them have low duty cycles, he/she can handle a lot more than 24
with a single T1 feed. Collisions will be rare, and for most
applications an occasional slight packet delay doesn't much matter.
If you do mpeg packet video, where a packet that arrives a millisecond
too late is as bad as one that never arrives at all, you start buying
Committed Information Rate, and you discover that 56K CIR is a lot
more expensive than 56K "bandwidth". You and your provider start doing
a lot of talking about and measuring of your data stream statistics.
If this has any relevance to the CIX, I've lost track of it.
--
Dick St.Peters, Gatekeeper, The Pearly Gateway; currently at:
GE Corporate R&D, Schenectady, NY stpeters@dawn.crd.ge.com
