& cplSiteName &
Comments
Newest First  |  Oldest First  |  Threaded View        ADD A COMMENT
<<   <   Page 3 / 10   >   >>
Rick.H
Rick.H
12/5/2012 | 2:45:08 AM
re: Nortel/Avici: Getting Together?
After securing its Next-Gen Edge router with "Neptune", Nortel is now seeking for a Core IP HW.
Neptune is based on brand new technology and should be shipping mid 2004. With these both products in hand, Nortel would make a nice come-back on IP. Definitely 2004 will be a great year...
Rick.
gotman
gotman
12/5/2012 | 2:45:08 AM
re: Nortel/Avici: Getting Together?
I wonder what other partners of Juniper are saying now!!! Lucent must be happy about this.. Don't know about Ericsson?

I think Nortel would have got more milage out of the partnership with Juniper!!!

Does anyone really think Avici has a roadmap beyond the exciting family of routers?

gm
Tony Li
Tony Li
12/5/2012 | 2:45:07 AM
re: Nortel/Avici: Getting Together?
This is ridiculous battle ground, but how do I build a non-blocking Procket beyond 480Gbps without paying for many costly interconnect ports? Avici scales card-by-card into multiple Terabits using the same fabric built into the line card. I don't really care if I need to add ports into another rack as long as I don't need to cluster routers, paying for costly interconnect chassis or interfaces that don't make me money. Procket spent all that time cooking processors, but still can't believe that they don't have the correct solution.
---------------------

Well, that would seem to be a matter of definition. Architecturally, you have the choice between incorporating a multichassis switch fabric into each chassis, or making it external. If you make it internal, you penalize locations that are using a single chassis. You also make it more challenging to build a switch fabric that provides any-to-any bandwidth and continues to scale. If you choose an external fabric, then you can optimize the internal fabric for the single chassis case, which is by far the most common case, and not make customers pay for an external switch fabric that they may not use. This is mostly a pay-me-now vs. pay-me-later tradeoff. If you make your switching elements independent, then it is possible to design a fabric that can scale to some extreme sizes, provide any-to-any bandwidth, and make the cost of the fabric depend only on the implementation technology.

---------------------
As I build my really dense router and I want to reliably home my trunk ports onto multiple line cards, why do I need a 40Gig line card to add my extra port? I can't believe I needed to start with redundant 480Gbs switch fabrics.
----------------------

You don't need redundant switch fabrics unless you want to provide redundancy in the switching plane. You could, not unreasonably, run multiple line cards with a single switch fabric. In fact, since the lowest reliability components are the optics, some folks may feel comfortable with redundant ports on the same media adapter, or redundant media adapters on the same line card. Of course, a large fraction of folks will want redundant everything anyhow.

Tony
FatherTime
FatherTime
12/5/2012 | 2:45:06 AM
re: Nortel/Avici: Getting Together?
Two words... Toroidal Mesh.

The world's most complicated switching fabric has to pay dividends eventually.
stomper
stomper
12/5/2012 | 2:45:06 AM
re: Nortel/Avici: Getting Together?
Tony,

I believe some of the statements you make here are
centric to the particular architecture Procket has
chosen, and not to switch fabrics in general.

I agree that once you first make the choice to use
an "all line cards talk to a central switch card"
architecture (formally an indirect network), then
you are next faced with the choice of incorporating the bandwidth to scale beyond a single chasis - or instead using a "rear" port
to allow scaling using an external solution.

However, Hyperchip, Avici, and the late Pluris all
first chose architectures based on direct
networks, where each node incrementally adds
switch fabric bandwidth, and a multi-chasis
buildout is just a natural extension of the
fabric and not a multi-tier network.

To "clockme"'s point, the rear facing port must
minimally be the full bandwidth of the single
chasis to maintain bisectional bandwidth
requirements, and at least 2X that in the
redundancy case. I can only stipulate a really
big optical crossbar of 2Tbps link bandwidth
(96 X 10Gbps X 2) ?

-S
unet
unet
12/5/2012 | 2:45:05 AM
re: Nortel/Avici: Getting Together?
stomper,
can you explain what you mean by "bisectional bandwidth requirements"

In general I see re-visit of central vs distributed switch fabric discussion of couple of years ago. I think current generation technology make 480G in one chassis very cost effective vs realizing the same capacity using distributed switching. However, the same technologies (PAM4 or the NRZ variants and TFI-5/SPI) should help distributed switch fabrics scale even higher -only if there will be demand for such hughe capacity. Avici, Caspian, Photuris,BrightLink, would be in the distributed league I think.
dljvjbsl
dljvjbsl
12/5/2012 | 2:45:05 AM
re: Nortel/Avici: Getting Together?
Lightreadign wrote:

These days, with telecom equipment moving more rapidly toward packet-based networks, itGÇÖs become more important than ever for Nortel to have a good IP routing story



Nortel also uses a lot of microprocessors, memory chips, FPGAs, power supplies etc in their systems. The cost of these and their technological capabilities seem to me to matter much more to Nortel's bottom line than any routing story.

So the real question is, why does Nortel or any other telecom vendor need a routing strategy?
jim_smith
jim_smith
12/5/2012 | 2:45:02 AM
re: Nortel/Avici: Getting Together?
Two words... Toroidal Mesh.

The world's most complicated switching fabric has to pay dividends eventually.


Wasn't there a startup 3-4 years ago (it went belly up) that was making a scalable cross-connect based on exactly that technology?
coreghost
coreghost
12/5/2012 | 2:45:02 AM
re: Nortel/Avici: Getting Together?
So the real question is, why does Nortel or any other telecom vendor need a routing strategy?


They need it to get to a full set of products.
In doing big deals, the broader your product
reach, the better deals you can create for
customers.

The situation right now is that certain vendors
have to bring in competitors (cisco or juniper)
into large deals. The competitor then gets
leverage into that customer.

The natural state for the larger telecom companies
is to have a full product line. That means a
core and edge IP solution.
coreghost
coreghost
12/5/2012 | 2:45:01 AM
re: Nortel/Avici: Getting Together?
1. Avici has the scaling solution that HFR is supposed to solve for Cisco and Gibson is for Juniper. (What was Procket's scaling solution again? Oh that's right--they don't have one).


I'll tell you what Avici has been told repeatly.
Their scaling solution is obsolete. Colocation,
a long metal bar and all those little pins
are not something that anyone wants anymore.
What was ok several years ago is not ok now.



2. Avici has sold core routers to RBOCs in the US and Europe. Who has Procket sold to again?


I dont know who you are trying to fool, but
Avici has one remaining customer which is
ATT. And that contract is mature to say the
least.


3. Avici has a chance at making a profit. Procket could go under if they don't make any large sales.

Avici has about the same chance of making a
profit as procket does. All procket needs to
do is sign one major contract to be in a
better financial position than Avici is.

Avici has been searching for a real second
customer for years now. They have been through
everyone multiple times with no takers.




<<   <   Page 3 / 10   >   >>


Featured Video
Upcoming Live Events
November 5, 2019, London, England
November 7, 2019, London, UK
November 14, 2019, Maritim Hotel, Berlin
December 3-5, 2019, Vienna, Austria
December 3, 2019, New York, New York
March 16-18, 2020, Embassy Suites, Denver, Colorado
May 18-20, 2020, Irving Convention Center, Dallas, TX
All Upcoming Live Events