[1841] in Commercialization & Privatization of the Internet
Re: Understanding the decision to Build a BIG T-3
daemon@ATHENA.MIT.EDU (Hans-Werner Braun)
Thu Jan 2 19:50:20 1992
From: Hans-Werner Braun <hwb@upeksa.sdsc.edu>
To: cook@tmn.com (Gordon Cook)
Date: Thu, 2 Jan 92 16:49:19 PST
Cc: com-priv@psi.com
In-Reply-To: <9201012227.AA14699@tmn.com>; from "Gordon Cook" at Jan 1, 92 10:27 pm
Gordan:
Let me emphasize a bit on Jordan's point. I think you interpret too much into
the "was built" comment. The NSFNET had quite a few more "phases" than was
broadly visible. A new architecture of one of the phases (the T3 network)
allowed for much easier modifications to accomodate new sites than was possible
before ("was built"). Look, the NSFNET backbone started with a 56kbps
network years ago. A new T1 infrastructure was built and integrated in 1988
and the old 56kbps network was phased out. This initial T1 network created
a logical network of sub-T1 circuits by utilizing intelligent circuit switches
and achieved much richer connectivity that way at a logical level. However,
it resulted in people being unhappy with sub-T1 speeds. In 1989 a quite
major new phase started when we took out the circuit switches, replaced
RT interface cards and moved to non-multiplexed T1 channels. This was NOT
done by creating a new network, but by integrating changed hardware, software
and circuits into the existing network. While it was done with great care,
it is a very complex undertaking, if you want to minimize disruptions of
services. In the same way, ANY significant expansion of the T1 network could
have had impact on the overall topology planning, as long as we wanted to
retain certain criteria (like three hops max as a network diameter and about
three redundant T1 links per site). If NSF would have been in a situation
of being easily able to always expand the network, it would have resulted
in serious network planning efforts, given the T1 network architecture that
we used. This was one of the reasons to design the T3 network quite differently
and to propose that to NSF. The T3 design, besides robustness and other things
mentioned earlier, should allow for expansion with fewer headaches. Remember,
at that time we already talked with multiple agencies about direct connections
or possibly even close collaboration, and NSF had even an open solicitation
out where submitters could propose being part of the new T3 NSFNET network.
My understanding is that NSF received and reviewed quite a few responses.
In any case, my main point is that the T3 network architecture is much more
easily able to accomodate new requirements, including new sites, than the
T1 network was. I suspect that that's what Al Weis had in mind with his
comment. No need for rather strange interpretations.
By the way, there were many other phases, like technology (hardware) and
software replacements. The network was an evolving entity, and still is.
Hans-Werner
><<MESSAGE from>> Gordon Cook 01-JAN-92 22:27
> cook@tmn
> In his open letter to the Internet Community Al Weis said:
>
> "Unlike the T1, the T3 network was built to accommodate the needs of NSF-
> sponsored institutions plus other potential users, including commercial
> users. This additional capacity was built using ANS funds. Any
> competitive advantage gained by ANS is appropriate, considering the risks
> involved and the experience gained in building the T3 network."
>
> Al, Hans Werner, Erik, Jordan or anyone who knows what I am missing please
> educate me. For I am confused by Al's statement which implies that the
> T-3 network COULD have been built in such a way as to ONLY accomodate the
> needs of NSF sponsored institutions without the "additional capacity" that
> used ANS funds. How is this possible? I thought T-3 was T-3 was T-3.
> Hans Werner's and Jordan's explanations were helpful in elucidating other
> fuzzy areas of the T-3 earlier in the week and since I am not a techie,
> perhaps there is something like the topology of core nodes versus end
> nodes that is the answer. (For example would it be laying T-3 pipes in
> the same topology as the T-1 network rather than using the MCI national
> backbone?)
>
> Whatever the difference, apparently there was a choice: build a T-3 to
> serve just the needs of the academic community or build a larger
> T-3 that could be sold to "other potential users, including commercial
> users." A decision was made to build the larger T-3. Apparently it was
> made without consulting the community as a whole. I do not wish to pass
> judgement on this decision, but rather to try to help to identify
> precisely what is dividing the community in the hope that such knowledge
> will be useful for opening channels of focused communication and hopefully
> lead to understanding and perhaps constructive compromise. Is my
> reasoning sound?
