tjs at msc.edu
Tue Nov 8 08:48:55 UTC 1994
> From: karl at mcs.com (Karl Denninger)
> Subject: Re: ATM Utility
> To: boone at prep.net (Jon 'Iain' Boone)
> Date: Tue, 1 Nov 1994 15:41:52 -0600 (CST)
> Cc: avg at sprint.net, tjs at msc.edu, nanog at merit.edu, nap at hq.si.net
[A long time ago, I wrote:]
> > > >cost-effective in a number of applications today. In particular,
> > > >the cost of wide-area DS-3 ATM services can be very attractive
> > > >when compared to a number of point-to-point DS-3s.
> > But, if you don't need the full 45 Mb/s, you can find a more
> > cost-effective solution in the wide-area Fast-packet services. In the
> > case of the MCI Hyperstream offerings, you don't have to pay for the
> > full amount of a circuit from point A to point B -- you simply pay a
> > monthly subscription fee and then a usage charge per Megabyte of data.
> > So, you can build a multi-megabit/s backbone that is (say) 10 Mb/s and
> > not end up having to purchase the entirety of the DS3 circuits needed to
> > provision it.
> Tell you what -- go run the numbers for any reasonable-sized IP provider,
> and tell me whether or not they are better off on "metered" service of this
> type, or with full-time dedicated circuits.
> Metered service will *always* be more expensive at reasonable to high loads,
> because the metering and billing costs money to do!
Right answer, maybe, but wrong question.
Most ATM pricing I have seen has a "committed information rate"
component, which assures a minimum available bandwidth. This is not
the "metered" pricing to which you are responding.
More information about the NANOG