AT&T Sets Copper Ethernet Course

Sources say AT&T Inc. (NYSE: T) is seriously considering VDSL-based technology for future Ethernet-over-copper deployments, a decision that could shift the balance among copper Ethernet startups.

What's interesting about the VDSL choice is that only one vendor is supplying that technology: Aktino Inc. Whether the startup has enough time to exploit that uniqueness is uncertain, though.

Other Ethernet-over-copper vendors support 2Base-TL, which is spelled out in the Institute of Electrical and Electronics Engineers Inc. (IEEE) 802.3ah standard for Ethernet in the First Mile and runs on the G.shdsl version of DSL. (See Aktino Dives Into Copper.)

Those vendors include two companies with ties to AT&T: Adtran Inc. (Nasdaq: ADTN), which won an Ethernet-over-copper request for proposals (RFP) in January, and Hatteras Networks Inc. , which last year scored a deal with BellSouth, now part of AT&T. (See Adtran Scores at AT&T and Copper Ethernet Snares an RBOC.)

Neither Adtran nor Hatteras seems in any immediate danger from VDSL, though. Michael Howard, principal analyst with Infonetics Research Inc. , notes that, while AT&T is believed to have VDSL-based Ethernet over copper in its future roadmap, "that's going to take a while to get to standards," with the first version not expected until the second quarter of 2008.

Aktino and AT&T declined to comment for this story.

Why does any of this matter, since Adtran presumably won the job? Because AT&T's choice could influence other carriers, changing the outlook for copper Ethernet. "If AT&T says they're going in some direction, I'm sure other people will follow," Howard says.

And the RFP that Adtran won reads as if the carrier has seen and liked Aktino's technology.

According to one source, who requested anonymity, the RFP says AT&T is interested in VDSL using discrete multitone (DMT) modulation and, to mitigate interference, multiple-input/multiple-output (MIMO) techniques. That happens to be the technology Aktino chose to pursue when it joined the Ethernet-over-copper field in 2003, well behind rivals such as Actelis Networks Inc. and Hatteras.

Aktino might not be alone in the VDSL camp for long, though. Actelis has VDSL-based gear in its plans, says Craig Easley, the company's associate vice president of marketing. "That's everybody's long-term direction," he says.

"There's time for all the players in this field. You can be sure Hatteras is going to protect its position as a leader in this field," says Chris Cook, Hatteras senior vice president of sales.

It's also worth noting that AT&T is a big place, and one RFP doesn't necessarily set the direction for the entire company. One source requesting anonymity notes the carrier has two outstanding RFPs for Ethernet over copper, one of which deals with long-reach and international cases that are well beyond the scope of the Adtran RFP, which covered U.S. territory only.

In any event, Aktino's position has gotten it noticed, as our first industry source says the startup has held recent meetings with major OEMs including Cisco Systems Inc. (Nasdaq: CSCO), which doesn't have any Ethernet-over-copper gear.

Cisco wouldn't comment on that, but the company says it's generally not interested in an Ethernet-over-copper acquisition. "Right now, we're sticking with our partnership strategy," says Mike Capuano, a Cisco senior marketing manager.

Hatteras is a part of that strategy. For about a year, Hatteras has been selling gear alongside a Cisco group that sells to non-ILEC clients such as Bay Area service provider Telekenex . It's not a formal partnership; rather, that Cisco unit refers interested customers to Hatteras. (See Telekenex Picks Hatteras.)

While Adtran isn't even confirming the AT&T win, it seems likely the company's Total Access 5000 would be the product line applied to the deal. The box isn't shipping in volume but has been in trials with a dozen customers, about a third of which came through with purchase orders late in 2006, says Kevin Morgan, Adtran's director of marketing for carrier networks. (See Adtran Touts Total Access.)

Competitors have questioned whether Adtran even has the technology to deliver Ethernet over copper, especially considering its Website doesn't list such modules for the Total Access 5000. But Morgan says Adtran's got the goods, including technologies like bonded copper, which combines copper Ethernet lines to boost bandwidth.

— Craig Matsumoto, West Coast Editor, Light Reading

Page 1 / 3   >   >>
stephencooke 12/5/2012 | 3:12:12 PM
re: AT&T Sets Copper Ethernet Course Seven,

I must apologize for not being clear. Perhaps a better description might be the following:

Where the distance can be made to be short (ie: between houses hanging off a pedestal) use DSL-based rings. VDSL2 gives sufficient bandwidth to make this approach interesting. There are other advantages, some of which I have mentioned in the earlier posts.

Where the distance is relatively long for DSL, use larger numbers of pairs with G.Bond to achieve sufficient bandwidth. Many DSLAMs already support bonding. I believe that T-Com was looking into offering a business service based on bonding.

Significant bandwidth, using existing Unshielded Twisted Pair (UTP). Seems to fit most of the points on one of my previous posts...

Thanks again,

stephencooke 12/5/2012 | 3:12:17 PM
re: AT&T Sets Copper Ethernet Course Hi Seven,

I never said that I would be using G.Bond on a single pair. Each house, serviced by the pedestal, has 2 pairs from the pedestal at a minimum; rings does not apply in neighbourhoods where there is physically only a single pair/house from the pedestal. If a single binder group is routed to the pedestal or if other binder groups at the pedestal are physically routed to different DSLAMs then my max bandwidth to that pedestal, and therefore around the ring, is 50Mb/s symmetric. I'm OK with that. As a consumer, access to 50Mb/s is much better than 2Mb/s, even if that 2Mb/s is dedicated.

Sharing the bandwidth is NOT the same as "3M/s to each home on average", especially given the QoS and multicast capabilities. P2P traffic is lower priority therefore it waits until there is bandwidth available that is not being used by higher priority traffic.

In terms of DSL bandwidth being limited by distance is the whole reason for doing this. Long haul optics need OpAmps and/or repeaters. Optical rings can be many times longer than individual span limits because they terminate & regenerate. Same principle here, it's do-able just tougher.

The high rates are on the individual links between houses, which ARE very short. In this case only 2 pairs are used for each house, one facing each neighbour (or the network gateway node on the ends). The link from the gateway node to the CO logically combines the pairs of all the subscribers on the ring (eg: if there are 8 subscribers then potentially 16 pairs can be combined towards the CO). This link is governed by the physics of DSL on each pair, the total available bandwidth is just multiplied by the number of pairs used.

If a telco, such as AT&T, can provide 50Mb/s (an increase of 25X over their current xDSL offering while simultaneously providing efficient multicast & QoS) without spending $??Billion to trench fiber, perhaps they can offer LightSpeed services to far more homes then their existing plans.

Incidentally, on links less than 1km (3.3kft) it is possible to get 400Mb/s with 4 people on the ring (or 1.5km - 5kft - with 8 people and over 2km - 6.5kft - with 12 people) so my statement holds in much of the US and almost everywhere in Europe depending on the number of people on the ring.

Thanks again,

paolo.franzoi 12/5/2012 | 3:12:18 PM
re: AT&T Sets Copper Ethernet Course

G.bond does not allow a single pair to run at 32x the speed of a pair without G.bond. What it allows is a binder group to run at (in theory) 32x the speed instead of 25x the speed. See the point? You are getting 32x instead of 25x. Or about 3 - 2. Thus theoretically, the 2Mb/s per pair becomes 3Mb/s per pair. So, now you have shared this across all these homes. It is no different than having 3 Mb/s to each home on average.

You can then provide a 1 Tb/s connection pipe at each home then throttled through the 3Mb/s per home connection.

So, by your latest claim you get 60Mb/s per 25 homes. Why is that any better than 3 Mb/s per home? It is as I said when web surfing. But as we move to machine to machine interfaces (like P2P) it is no different.

You are producing EXACTLY the same bandwidth as DSL. You are choosing to share it in a different way.

By the way, I am a DSL person as well. You don't seem to get the point that the bandwidth available via DSL is limited by distance. To get the high rates, the distances must be VERY short. Putting a ring in place is putting a repeater at every home. The point I am making is that to enter that network you have a long haul between entry to that ring and the Central Office. On that link between the start of the ring and the Central Office, the laws of physics must be obeyed. Therefore, you are limited by the distance that the ring starts. If that distance is significant (like a 12Kft CSA), then the DSL technologies are exactly what they are. You somehow think G.bond gives 32:1 gain in bandwidth per pair. That is simply a mistake. At best, it gives a 3:2 gain (Its actually less than that but I am rounding in your favor). So, again - you are going to have to run fiber from the start of your ring (12Kft in some cases) to the Central Office. I claim that this is the same as FTTC. Exactly the same. FTTC exists and 1 carrier deploys it broadly (AT&T SE).

Now you are claiming that having a 60Mb/s link from the neighborhood is good enough. At that point, you have gone from a claim of 400Mb/s shared to 60Mb/s shared. Which is what I have been pointing at.

Now, AT&T as part of Project Lightspeed is supposedly providing at least 20Mb/s unshared. In the case you have given, this is an equivalent of 500Mb/s and this requires less neighborhood construction than FTTP. Which is why AT&T has chosen this as their brownfield technology.

So, again - I am not hitting your local neighborhood wiring which is where your ring is. I am hitting how you connect your ring to the Central Office.

stephencooke 12/5/2012 | 3:12:19 PM
re: AT&T Sets Copper Ethernet Course DSL works today, in your situation at 12kft (3.6km) at say 2Mb/s within the existing binder groups (25 UTPs wrapped by a cotton ribbon). G.Bond will allow for 32X that bandwidth number less some overhead; the electronics, spectral efficiencies, NEXT, FEXT don't change. Even if the answer is 60Mb/s symmetrical that is a significant enough bandwidth increase in most cases. Note that this will also include QoS and efficient multicast.

This is a solution that brings significant service delivery bandwidth and all the trappings of a traffic-efficient system over the existing copper plant. The bandwidth offered is far better than straight DSL and, in many cases, almost as good as PON-based systems without the backhoe. An evolution path exists for these DSL rings to hang off of optical RPR collector rings, which is the same architecture as today's metro (ie: already very familiar to carriers).

The bandwidth that I will claim is what the VDSL2 chip suppliers can provide in a given situation. My available bandwidth number will have to do with a factor of 2X, as there are 2 directions around the ring, and nX, where n is the number of subscribers on the ring (max 16) up to 2X the max VDSL2 bandwidth (>200Mb/s or 100Mb/s symmetrical). It will also have to do with the distance that the pedestal is from the CO and the distance that the consumer is from the pedestal. Yes, 400Mb/s or 200Mb/s symmetrical is definitely possible in many cases. As a consumer, wouldn't you rather have bandwidth like this in a few years or are you willing to wait for a few decades for fiber to be installed, if it ever will be.

As an aside, implementing DSL Rings in an area will create a rather impressive barrier-to-entry for cable or other competitors at a relatively low installation cost point.

As a PON salesman I can see this technology making things more difficult for you, sorry. However, I came up with DSL Rings to help the carriers, not PON equipment providers.

All the best,

paolo.franzoi 12/5/2012 | 3:12:20 PM
re: AT&T Sets Copper Ethernet Course
There are many people who have worked on the DSL technology and DSM across binder groups. You do not get 5x the bandwidth that you can normally get.

32 pairs x 2M/pair = 64Mb/s not any more. That 2M per pair is an approximate limit of DSL technology with a full binder group at that reach.

So, again - Whatever you want to claim at the local house to house connection. But you would need about 200 pairs to each neighborhood, for those neighborhoods that would be DLC served. That is why people are doing fiber to those points. That is why FTTN and FTTC use Fiber in the F1 plant. That is why FTTC uses Fiber in the F2 plant.

paolo.franzoi 12/5/2012 | 3:12:20 PM
re: AT&T Sets Copper Ethernet Course
80 Mb/s at 500 feet. At 12Kft, all DSLs are created equal at a couple of Mb/s. So, you can not have high rate, long reach copper.


stephencooke 12/5/2012 | 3:12:20 PM
re: AT&T Sets Copper Ethernet Course 500ft is the loop from the pedestal to the house. The VDSL2 bandwidth in this part of the link doesn't change (eg: 320Mb/s or 160Mb/s symmetrical in the Bell South case). The link from the pedestal to the CO is G.Bond which has a maximum pair contribution of 32 pairs. At 12kft we can still get up to 32X the current DSL bandwidth depending on the number of subscribers in the ring. This is more than sufficient to make carriers think a few times more than they already do before deploying fiber.

stephencooke 12/5/2012 | 3:12:20 PM
re: AT&T Sets Copper Ethernet Course Seven,

No worries about criticality, if I couldn't handle it the idea would be dead already.

From the pedestal to the CO the link is G.Bond-over-DSL. The pairs that each subscribing household uses today from the pedestal to the CO are logically bonded into a single link of n X xDSL at the pedestal distance from the CO. I understood that the Bell South FTTC with VDSL2 was 80Mb/s symmetric. In the language of DSL chip vendors this is 160Mb/s as that total bandwidth can also be asymetric in nature (more downstream than upstream). This would yield 320Mb/s rings.

RPR is used in the ring providing protection switching, buffer management, efficient multicast and QoS by traffic type (implies deep packet inspection in the gateway nodes). SLAs based on traffic priority level are now possible. Equipment is deployed only as consumers request it (ie: Demand-based pay-as-you-grow), not the build-it-and-they-will-come of the FTTx world. This means that capital outlay is minimal up front. A feature of rings is that you can add/delete nodes without losing service to the remainder of the homes on the ring. Basically you add subscribers until the max bandwidth on the ring is close to that of the bonded link. If there are still houses to be serviced by that pedestal you can add totally separate, "stacked" ring(s). Given the ADM characteristic of ring nodes they can also be configured in strings for rural areas.

Another interesting feature of our implementation is that the gateways are trickle powered by the CO to maintain E911 when the power fails.

Incidentally, the upgrade path is RPR-based optical collector rings to the pedestals (the same architecture as in today's metro). At this point the number of houses on each ring can be reduced, meaning more bandwidth available per house (depending on QoS-based SLA), yet the bandwidth to the network would theoretically be maintained.

If you would like more information I would be happy to discuss it offline ([email protected]).


paolo.franzoi 12/5/2012 | 3:12:21 PM
re: AT&T Sets Copper Ethernet Course
Your several hundred megabits of bandwidth are shared across many homes. The question is how many homes. If you do it in a local area (say the F3 box) then you have to run fiber to that box as in an FTTC setup. FTTC already exists and carriers have already evaluated that. Only one carrier (BellSouth) has done a lot of work with FTTC. In the FCC mandates around FTTC (500 feet to the residence), an 80Mb/s connection to the home is achievable with VDSL2. I would say that a point to point connection with this technology is roughly the same bandwidth as what you propose.

If you say, no I am going to locate the terminating electronics back at the F2 box (where the DLC and the cross connect are) then your distances are problematic. You used a distance of 2000 feet in your calculation, but many homes are more like 8 - 10Kft from the cross box. On top of that you need to follow the binder groups, as once these split up there is no way to put electronics to bridge them unless you were going to span power it and make it sealed. Without such a box (or actually a number of them), the distance from the F2 box to your first home on the ring could be large. Thus, you will end up putting a large number of extra boxes out there simply as repeaters. The bandwdith achieved is approximately the same as a GigE fed IP DSLAM.

So, you are correct that you have a high physical clock rate at the access port. But it is a shared media (which PON is as well), so you have to divide the numbers by the number of homes on a ring. The sharing mechanism is different and more efficient for Internet traffic. It is not more efficient for P2P and other machine to machine interactions - where all the bandwidth requirements are.

Small rings = FTTC. Big Rings = IP DSLAM. So, I am not sure why you think this is such an improved concept in terms of bandwidth.

I am not trying to be super critical, but those are the challenges I see in bandwidth. I think the ring presents another whole set of issues and leave that to the casual observer to note.

stephencooke 12/5/2012 | 3:12:22 PM
re: AT&T Sets Copper Ethernet Course There was no "definitive assertion", I said that it "can deliver 400Mb/s" (B was corrected). In fact that bandwidth will be quite common but the typical will be on the order of 300+ Mb/s.

As DSL Rings will cost less than 1/10th of FTTH installation and provide substantial bandwidth, the question is whether installing fiber is cost-justified at all, even with operational challenges.

Page 1 / 3   >   >>
Sign In