Ethernet Operators Face Up-Front Costs

Plenty of vendors and service providers are betting big money on gigabit Ethernet becoming the predominant protocol in access networks, on the basis of its ability to deliver big bandwidths with low-cost equipment.

But significant upfront costs in deploying gigabit Ethernet -- laying fiber to the user’s door -- emerged at last week’s Gigabit Ethernet Conference in San Jose, Calif.

One service provider rolling out gig E to the home, the Grant County Public Utility District, said it would spend $120 million in the next five years, laying fiber to "all the farms, homes, and businesses" in the rural, mid-Washington region (see http://www.gcpud.org/zipp/default2.htm for details).

The county doesn't expect full payback on its investment for at least 15 years, according to Jonathan Moore, its senior telecommunications engineer.

"Right now we're spending $3,500 per home, but we expect that to drop to about $2,100 per home by 2003," he told delegates at the conference. The county already has 1,700 customers using the open-access network to order Internet, phone, and cable television services from a growing list of providers, paying approximately $75 to $80 a month for a combination of all three.

Other operators also appear prepared to pay big bills to connect customers, judging by comments made at the conference by Steve Albanese, director of information systems at CPI Wireless, a former division of Varian, which manufactures microwave and satellite components.

Albanese says his company's decision to purchase Ethernet services from Yipes Communications Inc. was helped in part by the service provider's willingness to dig a trench for free.

"We didn't have to put down a penny until it worked," says Albanese, who admits that CPI got "an early-user discount" when Yipes paid for about $40,000 worth of fiber-digging and hookup costs to link CPI Wireless's building with the local fiber loop. Yipes also threw in an Extreme Networks Inc. (Nasdaq: EXTR) customer-premises Ethernet switch as part of its service.

Free gifts aside, Albanese says Yipes's services delivered as promised, allowing the company to completely replace the frame-relay network it used to rely on for its 1,300-node network, which spans six locations in the U.S. and Europe.

"We now have four times the bandwidth for about half the cost of the frame-relay network," says Albanese, adding that provisioning headaches with the previous network of T1 (1.5 Mbit/s) leased lines "almost made me quit my job."

In general, the cost of laying fiber to customer sites "is not so exorbitant that it gets in our way," says Jerry Parrick, Yipes’s CEO.

Parrick points out that fiber-access alternatives to digging trenches do exist, such as existing power or sewage conduits. However, he admits that fully half of Yipes's current customers have needed some type of construction work to bring fiber inside, a task that Parrick says can cost between $50,000 and $60,000 per building. Yipes is sometimes able to share the construction costs with other service providers, he adds.

Of course, having to lay fiber can also delay service provisioning considerably, and might also limit the speed at which Ethernet-based services can be deployed in general. If you’d like to share your views on this topic, take Light Research's new interactive poll, at The Future of the Metro.

-- Paul Kapustka, Editor at Large, Light Reading http://www.lightreading.com

<<   <   Page 2 / 5   >   >>
kaps 12/4/2012 | 8:29:36 PM
re: Ethernet Operators Face Up-Front Costs Optinuts -- a bit of addl' info that got edited out of the main story may help you:

Grant County PUD has hyrdroelectric power plants (dams) on the Columbia River, and they resell excess power to places like California.

They're using some of this money to pay for the fiber buildout -- with the idea that two-way connections will let them better monitor power usage (so they know how much they can sell).

Between this application and their cut from providing services, they expect the infrastructure to be paid off in 15 years.

-paul k
John Honovich 12/4/2012 | 8:29:35 PM
re: Ethernet Operators Face Up-Front Costs Grant County's project is interesting but I can't fail to note that comparative solutions are considerably less expensive while generating similiar revenue.

Jonathan Moore is cited in the article as claiming projected revenue of $75 to $80 a month per customer for providing Internet, phone and cable services. The cost per customer is currently $3500 but will decrease to $2100 in 2 years. At this rate, the deployment is projected to break even in 15 years.

Consider this:

Marconi claims its FTTH DISCUS solution is $1100 per home.

Next Level's VDSL solution claims its payback is 6 years.

Let's grant that Ethernet is a more viable long-term solution. But at what cost.

With these numbers, World Wide Packet's solution is anywhere from 2x to 3x expensive than comparable converged internet, telephone and voice solutions. Given that they want ethernet to deliver cable TV, I also wonder about Grant County's need to install IP set-top boxes at a few hundred apiece for each TV. Is this cost included?

That being said, I agree with the sentiments of LightBeating. When the bandwidth comes, new applications will certainly arise. The question still remains is it worth the extra cost this solution incurs.

For example, LightBeating mentions that for $2000, he would seriously consider paying $12 a month for 15 years. But the true economic cost must include the cost of capital. Make this $2000 an annuity payable for 15 years at 10% interest, and the cost increases to $20 a month.

On a grander scale, you could make the argument for Grant County. Use a less costly solution (such as Marconi)and the $120 M cost may decrease to $60 M (depending, of course, on the percentage of costs the optoelectronics counts for).

The foregone interest payments in the cheaper solution alone could save millions.

On the other hand, how much more in increased revenues or operational cost savings does grant county have to achieve to justify their investment in GigE? Will their operating costs decrease by 30%? Will the revenues for Internet access be 50% greater? It would seem that their must be some significant changes in operational revenues and cost for this to be viable. And, if higher revenues, (in contrast to the Palo Alto FTTH project) it would strike me as peculiar that residents of Grant county would be particularly price-insensitive to paying for higher bandwidth.

Finally, why not delay. A 15 yr payback period puts them at great risk of being surpased by future disruptive technologies. It's not merely the fiber, it's the equipment attached to it. Surely 10 years from now, tremendous advances will be made in first mile optoelectronics. Does this not put Grant County at a long-term strategic disadvantage? Perhaps this wouldn't matter because the barriers of entry will be too great for multiple fiber networks in rural Washington, but it does not seem to be a prudent plan for competitive markets.

I'd be very interested in hearing everyone's thoughts.
gladysnight 12/4/2012 | 8:29:29 PM
re: Ethernet Operators Face Up-Front Costs LightBetaing wrote:
" . . . Those who dare lay fiber now are sure to be the winners."

I think you're almost right, but I think it's more a case of those who can survive the process of laying fibre now are sure to (eventually) be winners . . . .
Fred Snarff 12/4/2012 | 8:29:28 PM
re: Ethernet Operators Face Up-Front Costs I would be thinking twice about these business plans as well. HALF of all customers need construction?? Wow. That's a big red flag. And the people they are signing right up now should be the low hanging fruit. $50-$60K plus to dig, trench, lay conduit, pull fiber, then light it. Construction costs can easily go much higher than that too. You're talking MINIMUM 90 days and more likely 6+ months to provision a customer. All for something like $2000/month. I hope the E-LECs (ethernet LECs) have a ton of cash in the bank.

There is only one type of customer that will go through that pain to get connectivity - a big one. Large corporations with a dedicated IT staff and lots of time to kill while waiting for that nightmare process.

And I don't even know why the people on this board are talking about this stuff to the home. Are you serious?! Never. That's science fiction. The only place that's happening now (or any time in the near future) is in some municipal-subsidized science project. I like the guy who spends $10K on his home electronics and thinks he represents some significant portion of the general population.


fk 12/4/2012 | 8:29:24 PM
re: Ethernet Operators Face Up-Front Costs I agree with you that FTTH is largely a pipe dream right now, but never is a long time. It is conceivable that in a decade or two we will see FTTH, but it's definitely not a near term thing. The people who are going to be willing to pay for the sorts of services that make laying fiber worthwhile in the near terms are businesses and possibly large concentrations of residents (like apartment buildings).
DKP 12/4/2012 | 8:29:21 PM
re: Ethernet Operators Face Up-Front Costs
A few comments:

1. This is a pt-to-pt Gigabit Ethernet (GBE) Architecture. Currently GBE is still expensive. But watch the costs curves! GBE is only 3 years old. Don't bet against the price of Ethernet optics.

2. If you look from the 1970's to today, at no point did we ever anticipate correctly bandwidth demand. We have historically underestimated this.

3. Grant PUD is price insensitive, in a sense (ha). I wish that I were a utility re-seller right now;)

4. Here is an analogy for you. Coax penetrates 90%+ of homes in the US. Did that exists when you were in high-school? Nope. The entire country, all homes, were re-wired for CATV coax. What makes you think that fiber is more difficult? You lived through this! It happened in your lifetime, and it will happen again.

5. This architecture is pt-to-pt. A pt-to-multipoint passive gigabit Ethernet architecture is less expensive. The FTTC electronic switch box is $25K to $100K; that is one reason why the cost per user is high.

6. WWP has a good vision. They are betting on optical ethernet, and that is the best bet around right now. I think LightReading perhaps does not appreciate that FTTH trials like this do not reflect systems costs for volume deployments. When you talk about FTTH, you talk potentially about several million in volumes, and prices come down accordingly. FTTH architectures are just about on cost par with HFC, PacketCable, VDSL, etc right now. Check any other FTTH network solution - don't flag a GBE to the home as reflecting all FTTH architectures.

doco 12/4/2012 | 8:29:21 PM
re: Ethernet Operators Face Up-Front Costs > Jonathan Moore is cited in the article
>[....] The cost per customer is currently
> $3500 but will decrease to $2100 in 2 years.
> Marconi claims its FTTH DISCUS solution
> is $1100 per home.
> Next Level's VDSL solution claims its
> payback is 6 years.

I don't think that we are comparing apples to apples here. I think that the Marconi number is for just the equipment to hang on either end of the fiber and doesn't count fiber costs, much less installation costs. Is that $3500 number including either of those?

I can see that the cost of the raw fiber in the ground or strung on poles is about equal to raw copper today. The cost difference for FTTH (be it PON or Ethernet) is not the cable, and it is only slightly more for the install costs. The cost difference is for the electronics that get put on either end.

I see those equipment costs coming closer and closer once you throw data into the mix. For an HFC network you now have all the 2 way amplifiers, the CMTS and Cable modems. Those aren't free by any means. For DSL you have the DSLAM with it's required ATM switches, and the DSL modems.

I think that FTTH won't happen for the short term in areas that have copper today. But, anywhere someone is laying new cable anyway (be it green field, or competitive overbuild) the cost premium for fiber isn't that great and you can get more revenue for longer time period using it.

doco 12/4/2012 | 8:29:19 PM
re: Ethernet Operators Face Up-Front Costs > > With ethernet, the boxes are cheap, the network
> > expertise is (relatively) minimal,

> This is true as long as you use ethernet bridges in the core, which is
> hardly possible, because L2 networks do not scale up at all.

I would say that Ethernet ports (not just boxes) are relatively cheap. Look at the cost of GigE ports for a router vs. OC-12 ports. Or compare the cost of a couple of 100mb ports to OC-3 ports. You are talking a 10x price difference here.

I don't think Ethernet is being looked at as a wide/metro area network. It might be a link layer technology for those networks, but IP is being run over that to provide the network layer services.

netskeptic 12/4/2012 | 8:29:15 PM
re: Ethernet Operators Face Up-Front Costs > I don't think Ethernet is being looked at as a
> wide/metro area network. It might be a link
> layer technology for those networks, but IP is
> being run over that to provide the network
> layer services.

So, the dest MAC does not matter, source MAC does not matter and protocol id does not matter either, why would you (1) still call the thing Ethernet, (2) waste 14 bytes ?

I suppose that the only reason is marketing hype, which IMHO makes a lot of alarm bells to ring.

Just call the thing CFP (Cheap Fiber Packets) and then we can discuss how much will it cost and are there any benefits in adding STS-N structure to it.


DKP 12/4/2012 | 8:29:14 PM
re: Ethernet Operators Face Up-Front Costs > I don't think Ethernet is being looked at as a
> wide/metro area network.

> So, the dest MAC does not matter, source MAC
> does not matter and protocol id does not matter
> either, why would you (1) still call the thing
> Ethernet, (2) waste 14 bytes ?

IP is the converging layer(3), but a thin Ethernet layer (2) is still important. The one thing Ethernet does really well is transport packets. It provides 802.1Q (VLAN), 802.1p, 802.1w, etc features that IP alone does not have. You need to keep a thin Layer 2, either Ethernet or SONET, until these features are built into IP or the optical layer.
<<   <   Page 2 / 5   >   >>
Sign In