Anyone thought about eliminating large physical exchange points and replacing
them with a more distributed architecture?
Multiple data centers interconnected over ATM in a single metro area run by
indepdenant entities who are free to provide any level of service or value
add they wish.
Thoughts?
Anyone thought about eliminating large physical exchange points and replacing
them with a more distributed architecture?
Uhhh... Savvis.
Multiple data centers interconnected over ATM in a single metro area
That's the PacBell NAP in the Bay Area and the AADS NAP in Chicago.
run by
indepdenant entities who are free to provide any level of service or value
add they wish.
Kind of defeats the purpose of having an exchange point don't you think?
Or are you thinking of something more like what the PAIX does?
There is power point presentation floating out there about this.
-Madlion
Yes. Actaully there has been a bit of hypothetical conversation about
this very topic chez moi. I suppose this could be set up as a sort of
national extension of the CIX. The question that needs to be answered if
such a national ATM "exchange" were to be created is the issue of what
exactly it could be used for. Do you set it up like a national exchange,
and only allow providers to exchange traffic with other providers, or do
you set it up more like the Ebone, allowing providers to connect in
multiple metropolitan areas and use the national exchange's bandwidth
instead of building their own backbone? And do you set up a national ATM
cloud connecting all the existing NAPs and allow anyone to peer with
anyone else anywhere, or simply have a national ATM cloud directly
connecting providers, or both? Of course, it would be a bit harder to get
all the FDDI NAPs onto ATM but isn't WorldCom supposed to have ATM at
MAE-East by Q3 this year (leaving only PAIX and SprintNAP)? At this
point, sending packets via FedEx seems to be more reliable than MAE-East,
(FedEx is a little slower but at least ALL the packets get there
eventually) so any functional alternative is welcome...
-Blake
black@bleep.ishiboo.com wrote:
Anyone thought about eliminating large physical exchange points and replacing
them with a more distributed architecture?Multiple data centers interconnected over ATM in a single metro area run by
indepdenant entities who are free to provide any level of service or value
add they wish.Thoughts?
It sounds alot like the multi-tiered proposal I put forth about 3 years
ago to build local exchanges on top of frame networks in certain cities,
then interconnect cities to regional exchanges, then interconnect the
regional exchanges. That architecture takes advantage of traffic
locality as well as providing a path out for non-local traffic. If the
individual regional exchanges have a small enough number of
participants, they are easier to manage, and should one participant have
alot of traffic going to regional or inter-regional exchange, you simply
install a PVC to there. There is some breakpoint for scaling however...
I think it still would work pretty well on top of Frame, though it is
aimed at the local/regional ISP rather than the nationals.
Bob
> Multiple data centers interconnected over ATM in a single metro area
That's the PacBell NAP in the Bay Area and the AADS NAP in Chicago.
Don't forget the IndyX, and the VNAP architecture.
A little bit of both (Regional / National).
And, Priori is specializing in setting up regional
exchanges.
Quite a few come to mind....
Subject: a little thought on exchanging traffic
Anyone thought about eliminating large physical exchange points and
replacing
them with a more distributed architecture?
Actually, this is a good idea, if the co-location or adequate locations to
place access equipment and transit routers can be found in the area, or
build-out for each provider is an option.
If I could charge for the cross-metro transit at say, $1000/month for a
dedicated T3, do you think that the concept of a carrier hotel would go
away? If so, which cities should be first (like toasting MAE-EAST?)?
On a similiar note, what if a company were to create a "virtual exchange
point", where backbone providers could order service anywhere in the country
and get transit to their peers elsewhere for a flat rate. IE, if I run a
4xOC-12c ATM network nationwide, and let transit occur between peers over it
rather than at a co-lo facility. I can imagine charging around 5K a month
for a DS3 circuit with PVC's to around 5 peers.
Opinions anyone? (VC anyone?)
Jonathan Arneault
Director of Business Development
CMA / INet Solutions Division
518.783.9003 x 253
Naaaah, Bob. That would mean that there might actually be some
geographic locality of reference to Internet traffic -- you know, my
telnet from Tampa to Auburndale, Florida, might actually not go via
Orlando, Atlanta, DC, and MAE-East in New Jersey.
And that would never be acceptable.
Just ask Sprint, AT&T, GTE/BBN, and MCI.
Cheers,
-- jra
Whoa, your idea is not entirely new, just now coming into
vogue..
We put up a prototype in '95, it cut into production a few
months later.
This is a Regional Exchange with ATM access to the national
exchanges.
(Including switches at FDDI exchanges) We sell longhaul
paths,
IP transit in Indy, and local exchange services. It seems
fairly scalable....
Richard
Jonathan Arneault wrote: