[8915] in Commercialization & Privatization of the Internet
Re: How Will NSF Get Out of NSFnet Business? 3 Possible Scenarios.
daemon@ATHENA.MIT.EDU (Marvin Sirbu)
Sun Dec 12 12:03:47 1993
Date: Sun, 12 Dec 1993 12:01:40 -0500 (EST)
From: Marvin Sirbu <ms6b+@andrew.cmu.edu>
To: com-priv@psi.com
Cc:
In-Reply-To: <9312111803.aa22008@pandora.sf.ca.us>
It's worth pointing out that the $18 million that NSF said it will spend
as part of its resolicitation is for all aspects of the new scheme
- the VBNS
- the Routing Arbiter
- the NAPS
- four year phaseout of funding to Regionals
No new "backbone provider" is going to get $18 million.
Furthermore, there are only three priority NAPS in the solicitation. If
the NSF decides to fund only three, they will take on a character more
akin to the CIX than to points of attachment to a backbone. Indeed,
most cross country transport will be provided by whoever carries traffic
from campuses to a NAP. In that sense, there is no longer a "backbone".
For example, all traffic between institutions west of the
Missisippi--e.g. a packet from Dallas to Seattle--will go via a
combination of regional and commerical NSPs to the west coast NAP, and
then out again via a commercial NSP and regional to the destination. No
backbone.
The vBNS will link only the supercomputer centers, so it cannot be
viewed as a substitute for the existing backbone. There is no provision
for random Universities to get access to the vBNS except via a NAP, or a
special grant for "meritorius connections"; if the packet has been
carried all the way to the NAP already, the VBNS is not going to have
much left to do. Thus, "backbone services", with the exception of the
supercomputer centers, under the new arrangement, will be subsidized by
the NSF only to the extent of grants to the Regionals which can be used
for purchasing access to NAPS via commercial Network Service Providers.
These subsidies are scheduled to be phased out.
In no way should the vBNS be viewed as simply a higher speed replacement
for the existing backbone. The bulk of the production traffic in
support of R&E will have to go over non-subsidized commercial Network
Service Providers.
Here's a different scenario:
The NAP and Routing Arbiter contracts are awarded. A regional like
Prepnet which is not near a NAP contracts with a commercial NSP, say
Sprintlink, to carry traffic from Pennsylvania to the New York priority
NAP. NearNet decides to lease its own T3 from Boston to New York to
provide its connection to the NAP. Packets from Penn State to MIT go
via Prepnet, Sprintlink, the NAP, the leased line and NearNet.
Barrnet connects directly to the San Francisco NAP. Prepnet and BARNet
pay PSI for inter-NAP connectivity. Traffic from Penn State to Stanford
goes from Prepnet to the NY NAP via Sprintlink, then via PSI to the SF
NAP and then to Barnet.
Prepnet, Barnet and Nearnet get grants from the NSF to subsidize their
connections to the NAP, and their payments to Sprintlink, PSI and
Nearnet's T3 supplier.
As the subsidies to the regionals decline, they raise their prices to
the campuses. Eventually, Sprintlink comes to Penn State and says, we
will carry your traffic directly to any of the NAPs for a fee comparable
or lower than what PrepNet is charging to cover its costs, plus payments
to Sprintlink and PSI. PSI, afraid of losing the inter-NAP business,
comes in with a lower price still. Penn State decides to attach
directly to PSI, and Prepnet starts to whither away.
Meanwhile, MFS goes to Nearnet and says, instead of
leasing a T3 to get to the New York NAP, buy an ATM virtual circuit from
us. Nearnet does so. Later, MFS persuades Nearnet to eliminate its
leased line network and simply interconnect the campus routers of all
its customers using ATM permanent virtual circuit connections to the MFS
ATM cloud. Finally, it persuades MIT that a direct attachment to its ATM
cloud can be used to provide virtual circuits from MIT to all three
priority NAPs. This saves MIT the cost of payments it was making to
Nearnet.
Since the New York NAP was being provided by an ATM switch supplied by
Nynex, MIT can now get a direct ATM virtual circuit via the New York NAP
and the vBNS to the San Diego Supercomputer Center for its visualization
work, obviating the use of IP level routers altogether. (This last
scenario is not likely to materialize for at least four years due to
wrangling over ATM network-to-network interface (NNI) standards.)
Marvin Sirbu
