x
Optical/IP

Atrica Keeps It Simple

A newcomer to the fiercely competitive optical Ethernet market may have found a way to distinguish itself among cash-strapped carrier customers.

Atrica Inc., born in February 2000 (see Out of Atrica), today launched its first products -- including two networking boxes designed to put legacy carrier traffic onto Ethernet in metropolitan area networks. Atrica says this will enable telecom carriers to consolidate both legacy voice and data traffic onto Ethernet links that are cheaper and easier to maintain (see Atrica Intros Optical Ethernet System).

Atrica also released the news that BellSouth Corp. (NYSE: BLS) has joined its board and will contribute an undisclosed amount to the startup's funding.

Atrica hopes to get an edge over rivals such as Appian Communications, Luminous Networks Inc., Tropic Networks Inc., and World Wide Packets Inc.

In general, Atrica's new products aim to deliver the economics of Ethernet without extras like synchronous voice, special video support, or other features that have typified announcements from some rivals.

Instead, Atrica stresses the basic value proposition of WAN Ethernet for less. It says its wares will sell for $1,200 to $3,500 per gigabit port, compared to metro Sonet prices of $18,000 per port for OC12 (622 Mbit/s) -- figures Atrica says it's taken from estimates by the The Dell'Oro Group.

"There's nothing fancy here. Atrica's doing flat-out Ethernet, competing strictly on price/performance," says Michael Kennedy, managing partner at Network Strategy Partners LLC, a consultancy.

Atrica's new products include the A-2100 Optical Ethernet Edge Switch, designed for the basements of customer premises or for carrier points of presence (POPs). The box takes in multiple TDM (time-division multiplexed) leased lines at rates from DS1 (1.544 Mbit/s) to T3 (45 Mbit/s) and uses circuit emulation to put them onto gigabit Ethernet connections in 1-Mbit/s increments. TDM voice is encapsulated in Ethernet using ITU spec G.744. The A-2100 also has OC3 (155 Mbit/s) or OC12 interfaces, so it can turn Sonet links into enterprise Ethernet bandwidth. Each A-2100 has a full-duplex capacity of 4 Gbit/s.

A second product, the A-5100, is designed for POPs of central offices. It supports up to 24 1-Gbit/s links and is designed to aggregate traffic from multiple A-2100s, supplying redundancy for all links. Both boxes are available now, along with an element management system that provisions services, monitors traffic for SLA (service-level agreement) verification, and uses Corba to tie into external OSSs (operation support systems) at carrier networks.

In many ways, Atrica's saving its best features for later. The A-8100, a 10-Gbit/s Ethernet core switch, is slated to ship during the third quarter of this year. The A-8100 will introduce MPLS (multiprotocol label switching) technology to Atrica's product line and will give carriers a way to replace OC3 and OC12 Sonet links at the metro core network.

And, at the Supercomm trade show in June, Atrica plans to announce a product that will support WDM (wavelength-division multiplexing), which will further extend the Ethernet capacities and distances Atrica's boxes support.

What is different is Atrica's stress on big carriers -- specifically, big international carriers interested in improving their transmission speeds with a minimum of fuss and cost. Atrica's enlisted these carriers as both funders and customers. In addition to BellSouth, Atrica's announced that four service providers, Bezeq, France Telecom SA, SBC Communications Inc. (NYSE: SBC), and Telia AB, have contributed $18 million to Atrica's $40 million in funding to date.

Customer-wise, trials are underway at Bezeq and France Telecom (see Atrica Announces Trials, Appointments). Atrica also says it's enlisted a German carrier and a trio of U.S. providers as trial customers. (The identities of these new customers are being kept under wraps, but sources indicate that one of them is Deutsche Telekom AG [NYSE: DT]. And it's likely that BellSouth and SBC would test Atrica's wares in the U.S.)

Atrica's focus on the big carriers is important. Given the current financial climate, experts say it will be these, not specialty providers, who ultimately finance a big buildout of Ethernet services (see Endless Ethernet?). And carriers are likely to help Atrica fund a product that will actually sell, instead of one that's developed at arm's length from customers.

The products are still in trial and have yet to ship, however, so time will tell whether Atrica's strategy is on the mark.

- Mary Jander, Senior Editor, Light Reading http://www.lightreading.com
<<   <   Page 2 / 3   >   >>
hujifan 12/4/2012 | 8:31:13 PM
re: Atrica Keeps It Simple You're wrong!

You're talking about 10GE LAN interfaces - these will be shipped to the enterprise market.

Everyone who's talking about Ethernet in the Metro are talking about 10GE WAN, which is identical to OC192c, except for the timing and synchronization module, which costs a couple of hundreds $ (actually, PMC sells the same framer for OC192c and for 10GE WAN). So if you look at the potential markets, OC192c has actually a larger market than 10GE WAN.

netskeptic 12/4/2012 | 8:31:12 PM
re: Atrica Keeps It Simple > Recently priced transponders show that an
> OC-192 transponder (includes laser, CDR,
> mux/demux, photodiode) is roughly $4,500 in
> volume purchases. 10GE transponders with the
> MAC chip can already be found for as low as
> $2,000. So, the OC-192 stuff is definately more
> expensive from a component standpoint.

I suppose we can optimize existing SONET chipsets for simple POS connections, e.g. without any SONET line and section services, just framing. It will drive price down, while preserving the possibility to populate board with more expensive parts and get full blown SONET if necessary. In my view this will beat 10 E as a value proposition by providing a more reasonable combination of price and flexibility.


> Moreover, when looked at from the system level,
> 10GE switches will be far cheaper on a price-
> per-port basis than OC-192 Crossconnects. This
> is due simply to the volume of shipment in the
> Ethernet world. Compare the volume of shipment
> of something like a Ciena Coredirector
> (probably measured in the hundreds) to an
> Ethernet switch like Cisco's Catalyst 6xxx
> (measured in the hundreds of thousands). The
> volume of shipment allows the vendors to
> amortorize the development and operational
> expenses over many more units.

Let us compare apples with apples again. If we have a packet switch in the middle, the particulars of framing used to deliver packets is not going to matter much.

Also, it seems to me that you are talking about true 10G ethernet switching here, which brings us back to the question of handling floods and spaning trees at 10Gps.


Thanks,

Netskeptic

cfaller 12/4/2012 | 8:31:11 PM
re: Atrica Keeps It Simple So 10GE won't be cheaper than OC-192? What's the bottom line? Does anyone know?
fiber_r_us 12/4/2012 | 8:31:10 PM
re: Atrica Keeps It Simple >You're wrong!

I'm not, usually! :-)

>You're talking about 10GE LAN interfaces - these
>will be shipped to the enterprise market.

Yes, I am talking about 10GE LAN (but the same chip supported the WIS function as well, so you could run it as WAN if you wished). In any case, the 10GE (WAN or LAN) tranponder is cheaper than the SONET OC-192 transponder.

>Everyone who's talking about Ethernet in the
>Metro are talking about 10GE WAN, which is
>identical to OC192c, except for the timing and
>synchronization module, which costs a couple of
>hundreds $ (actually, PMC sells the same framer
>for OC192c and for 10GE WAN).

Everyone?? I don't think so. The 10GE LAN solution will work just fine in the metro. As everyone has been pointing out, the optics and reaches are the same as a SONET OC-192 transponder. So, there is no problem using the LAN version directly on a fiber. The only reason to use the WAN version is to make it backwards compatible with the clock speed of an OC-192. The 10GE WAN solution is not "identical" to an OC-192:

1) It does not support protection switching.

2) It's clock tolerance in not within SONET specs (100ppm vs 20ppm).

3) It does not support other overhead bytes such as J0 and DCC.

The WAN standard is useful if you want to tranport Ethernet packets over an existing OC-192 wavelength. If you don't want to do this, then there is no purpose in using the WAN standard. Besides, you loose over 400Mb/sec of throughput in the WAN standard compared to the LAN standard. Why would you take this hit if it didn't need to be run over a SONET network?

>So if you look at
>the potential markets, OC192c has actually a
>larger market than 10GE WAN.

I don't know how you came to this conclusion. How is it that anything in the carrier space has a larger market than the Enterprise space? Even carriers have "enterprise" networks internally! The market for 10GE LAN is at least an order of magnitude greater than SONET, just like GigE is already greater than SONET.
fiber_r_us 12/4/2012 | 8:31:08 PM
re: Atrica Keeps It Simple >I suppose we can optimize existing SONET
>chipsets for simple POS connections, e.g.
>without any SONET line and section services,
>just framing. It will drive price down, while
>preserving the possibility to populate board
>with more expensive parts and get full blown
>SONET if necessary. In my view this will beat 10GE
>as a value proposition by providing a more
>reasonable combination of price and flexibility.

I don't think this is an issue of the complexity of SONET chipsets or Optics. It has everything to do with what the component vendors can get away with when it comes to pricing SONET compliant components vs Ethernet components. The competition is much more aggressive in the Ethernet component space than it is for SONET components.

Even if the prices for these specific components (i.e. transponders and MAC or SONET framers) were the same, these components only make up a small part of an overall system's cost. Lets take the Coredirector:

- It will have 32 OC-192 interfaces in one rack
rack
- Each OC-192 transponder component costs ~$4,000
- So, the total of all of the transponders costs ~$128,000

The overall system costs much more than this to build. You must include the switch fabrics, mangement and control systems, mechanical hardware, power supplies, software, non-recurring engineeering time, manufacturing costs, etc... Overall, the Coredirector (or any other BXC), must then spread this development expense, component cost, margins, and manufacturing expense over the number of systems they predict to sell. A BXC might sell a few thousand systems over a few years. In all, the component costs become a small issue in the overall product costs.

An Ethernet switch, on the other hand, may have the same magnitude of initial costs, but these costs are spread over what may be hundreds of thousands of units. The volume makes Ethernet a cheaper product to sell.

So, even if SONET components become cheaper than Ethernet, where does the volume come from? It is not likely that Enterprises will deploy SONET gear. Even the carriers do no use SONET gear for thier internal networks.

> Let us compare apples with apples again. If we
> have a packet switch in the middle, the
> particulars of framing used to deliver packets
> is not going to matter much

Agreed. The framing doesn't matter. So, why not pick the cheapest one and the one in which the frames were originated: Ethernet. No computing system I know of has SONET interfaces. I don't think you can even buy an OC-x interface for servers or workstations. If you can, I don't think that it would be price competitive or functionally as useful. 99+% of data traffic started out as Ethernet packets.

>Also, it seems to me that you are talking about
>true 10G ethernet switching here, which brings
>us back to the question of handling floods and
>spaning trees at 10Gps.

Floods and spanning tree would only be an issue if you used large numbers (100s) of Layer 2 switches meshed together. A Layer 3 switch (ie. a IP router or MPLS router) does not have these issues. I would never suggest building a large-scale layer 2 network. Such a Layer 2 network would suffer from all sorts of scaling problems and indeed is a bad idea.












netskeptic 12/4/2012 | 8:31:07 PM
re: Atrica Keeps It Simple > I don't think this is an issue of the
> complexity of SONET chipsets or Optics. It has
> everything to do with what the component
> vendors can get away with when it comes to
> pricing SONET compliant components vs Ethernet
> components. The competition is much more
> aggressive in the Ethernet component space than
> it is for SONET components.

So, the reason for savings is the lack of market presssure on SONET vendors :) - I suppose we are going to see changes here :).


> Floods and spanning tree would only be an issue
> if you used large numbers (100s) of Layer 2
> witches meshed together. A Layer 3 switch (ie.
> a IP router or MPLS router) does not have these
> issues. I would never suggest building a large-
> scale layer 2 network. Such a Layer 2 network
> would suffer from all sorts of scaling problems
> and indeed is a bad idea.

I suppose that any network utilizing 10 Gbps backbone is qualified to be called large.
So, the 10 GE links are supposed to be terminated in the router anyway. So, where this huge economy of ethernet equipment scale will come from ?

So far, I saw nothing contradiction the notion that 10 GE is anything else than hype trying to lure public dissatisfied with MPLS into the new-new-new thing.

Thanks,

Netskeptic

fiber_r_us 12/4/2012 | 8:30:59 PM
re: Atrica Keeps It Simple > So, the reason for savings is the lack of
> market presssure on SONET vendors :) -
> I suppose we are going to see changes here :).

No doubt there will be increased pressures for the SONET gear vendors to lower prices. But, the Ethernet guys have a lot more margin to play with. The SONET equipment vendors will be accelerating thier way to Chapter 11 if they try to compete with Ethernet on price.

>So, the 10 GE links are supposed to be terminated
>in the router anyway. So, where this huge economy
>of ethernet equipment scale will come from ?

The 10 GE routers are the same routers that the enterprise customers use. The enterprise customer drives the volume.

>So far, I saw nothing contradiction the notion
>that 10 GE is anything else than hype trying to
>lure public dissatisfied with MPLS into the
>new-new-new thing.

I don't see how you come to this conclusion. The idea that a better and cheaper network for the data world can be built; and that it may be based on Ethernet and MPLS can hardly be construed as "hype". Is everything that is new "hype"?


netskeptic 12/4/2012 | 8:30:58 PM
re: Atrica Keeps It Simple >> So, the 10 GE links are supposed to be
>> terminated in the router anyway. So, where
>> this huge economy of ethernet equipment scale
>> will come from ?

> The 10 GE routers are the same routers that the
> enterprise customers use. The enterprise
> customer drives the volume.

Are you trying to tell me that a router equipped with 10 G POS interface is a way more expensive than the same router equipped with 10 GE interface ?


> I don't see how you come to this conclusion.
> The idea that a better and cheaper network for
> the data world can be built;

I am sorry, I do not see theis 'better and cheaper' element.

> and that it may be
> based on Ethernet and MPLS can hardly be
> construed as "hype".

Ethernet does not add any value here, IMHO. The only function it serves is to provide frustrated public with yet another glimpse hope to get simple inexpensive backbone technology.

> Is everything that is
> new "hype"

So, far (say through 90s) 80% of all things new was either hype or something badly poisonned by hype. I would not be surprised to find out that MPLS is hype too.

Thanks,

Netskeptic

gladysnight 12/4/2012 | 8:30:55 PM
re: Atrica Keeps It Simple > The 10 GE routers are the same routers that the
> enterprise customers use. The enterprise
> customer drives the volume.

Are you trying to tell me that a router equipped with 10 G POS interface is a way more expensive than the same router equipped with 10 GE interface ?
======================

No, I don't think so. What he's trying to say is that 10GE (LAN or WAN but probably LAN) links will have enterprise routers at one end of the fibre span and telco switches/x-connects at the other. As such, there will be a lot more of them than there are internal (telco) sonet/sdh interfaces. Therefore, on average, they will be cheaper.

WAN-PHY (the 10GE wan spec) would seem to be dubious, given that as a telco you go to a lot of trouble/expense to replace existing sonet/sdh investments, and at the end of it you have:
1 - the same or possibly less capability than you started with
2 - the "benefit" of being able to say "Look! It's ALL Ethernet!"

Hoo-bloody-ray.
fiber_r_us 12/4/2012 | 8:30:52 PM
re: Atrica Keeps It Simple >Are you trying to tell me that a router equipped
>with 10 G POS interface is a way more expensive
>than the same router equipped with 10 GE interface ?

Remains to be seen if a 10GE interface on a M160 or GSR is cheaper than a OC-192 on the same box. Historically, Ethernet has been slightly cheaper, even on these platforms. The Ethernet interfaces have always given a better price/performance than the SONET interfaces. However, my point was that you will be able to buy routers, other than the M160 or GSR, that have 10GE interfaces (and may not even support OC-192) for a lot less than a M160 or GSR. Look at something like Cisco's 6500 or many of the startups.
<<   <   Page 2 / 3   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE