Are We to Buy Speed for a Few or Connectivity for Many?
dennis at MrBill.CAnet.CA
Sat Feb 1 21:54:56 UTC 1992
> While there maybe "bad" regionals that need to be hardened (sorry Laura,
> I can remember the new politically correct word, i'm not very good at
> this politics stuff), there is a much larger need to harden the backbone
> infrastructure through alternatives, the CIX being one. Its pretty
> frightening what "the market" could do with a 1/8th of the funds being poured
> into the current nsfnet. If you look at the upcoming solicitation from
> NSF they say that implicitly - you can do the math yourself.
This leads to a topic which I personally find interesting. One advantage
of a single, large (well-working) "backbone", with the asymmetric relationship
to client networks which this implies, is that it simplifies routing
considerably for just about everyone. Making two or more points of
contact work well between networks which wish to maintain some sort of
symmetric relationship makes the problem of obtaining decent routing
between them quite hard.
The two-backbone NSFnet proposal, and the CIX model as I understand it,
both seem to have the potential for producing some really intriguing routing
problems. The NSF seems to be covering their bets by also spending
big bucks for a routing arbiter, whose problem I assume this would become.
>From here it appears that the CIX has punted on the problem for the moment
by only bringing up one point of contact (I assume that packets wouldn't
travel between Reston, VA and Falls Church, VA via California if this
weren't the case). While this does simplify things considerably, it
can obviously lead to some very long distance paths between geographically
close sites and (speaking of hardening) leaves a single point of failure
where a fire or earthquake or backhoe may ruin a lot of peoples' days.
I was wondering, then, if there was a routing plan for a possible future
CIX with more players and more points of contact (or even a plan for
bringing up the rumoured CIX-east), and if so is the plan available.
In fact, I'd also like very much to see what the NSF had in mind for
the two backbones in terms of routing, though I suspect even they aren't
quite sure at this point.
More information about the NANOG