RE: too many routes

Mailman List Mirror (Read Only)
1

The routes issue historically comes down to the fact that Sprint did not
want to convert from Cisco 4000 to Ciscos that had larger memory capacity.

Sprint never used cisco 4000s in the backbone. Just FYI.

Historically, memory limitation was because CSC/4 board in AGS/+
routers had memory soldered in. The box was absolute top of the line
when it started to fall over.

Memory is cheap these days ... the big boys just don't wish to have a
free market.

This statement shows that the level of comprehension of the issues
remains absymally low.

It is NOT memory; it is CPU which is a limiting factor. Even the
mainframes would keel over on routing computations if the drastic
measures weren't taken to aggregate and dampen.

Now, can we stop spreading the "no memory" 5 year-old news?

Deny /19s and or a transition to IPNG then deny Peering to keep the market
from being open.

Hey folks, it is not closed. Keep the faith and let the big boy bleed market
share. I would hope that ARIN, RIPE, and APNIC would have the guts to keep
giving routable blocks to new contenders.

Oh, yeah. How clueful. Nowadays only a telco or an oil company
can afford to get into the backbone market. IP allocation is an
insignificant detail given the $100mil-to-get-leg-in-the-door of the
backbone market.

Please people, we must stop abstructions to keep the market open and
competitive.

Can you spell "economies of scale"? Or "using fiber at cost means
owning the fiber"? If you want to play the backbone game you've
got to own long-haul transmission facilities.

A small backbone provider simply cannot be competitive; no more than
neighbour garage can compete with Chrysler.

--vadim

2

Vadim Antonov <avg@pluris.com> writes:

Sprint never used cisco 4000s in the backbone. Just FYI.

True. The only one I ever remember was the DANTE access
router which was dropped in place before AS-path
prepending in software came about, and that wasn't a
backbone router by any stretch of the imagination.

There may have been one deployed to handle NACSIS or some
other odd case at some point, however I don't remember it
concretely.

Historically, memory limitation was because CSC/4 board in AGS/+
routers had memory soldered in. The box was absolute top of the line
when it started to fall over.

Sprint, in fact, was in an OK position since it was
possible to eliminate full routing on most of the AGS+es in
relatively short order once the memory problem became more
apparent. I believe you and I got to the point where none
of our AGS+es had full routing before people like JANET
started having regular problems and had to start dropping
routes from places like PSI entirely in order to have any
connectivity at all.

The AGS+ -> 7000 migration was a conceptual disaster,
since the RP+SP combination simply didn't perform well
under load.

The SSP made things happier, particularly when things like
selective packet drop were implemented, and this in
essence led to Peter's frequent (and pretty accurate)
complaint that the initial introduction of the 7500
compared to a 7000 + RP + SSP was a step backwards.

(Several people who used to be at cisco, and some who
still are, also have been heard making that complaint from
time to time...)

Can you spell "economies of scale"? Or "using fiber at cost means
owning the fiber"? If you want to play the backbone game you've
got to own long-haul transmission facilities.

And if you own such transmission facilities you have to
hope that your competitiveness police don't force you to
separate your transmissions facilities from your
value-added services, and require the transmissions people
to offer consistent pricing to all parties including the
value-added services people.

Of course, if bandwidth becomes much cheaper than you had
bargained for initially, it could be a disadvantage to be
trapped into using your own company's transmissions
facilities with a requirement of a decent ROI, so it kinda
cuts both ways.

  Sean.

3

Sorry to say "me too", but this is a point that is well worth reading
through twice, especially with todays fiber markets. With the major build
outs of many alternative long haul fiber companies (Qwest, IXC, etc etc),
it is becoming /very/ unclear that being a facilities based provider offers
you /any/ advantages whatsoever in the realm of price/performance. I would
argue that just the opposite is true.