Will the Internet lead fiber into the access network?
Of the several factors cited these days for the current fiber-optics boom, the advent of the Internet--and the expansion of data traffic of which it is a part--has developed as the most conspicuous. Bandwidth demands, say the pundits, show no sign of slowing their rapacious growth as data threatens to overtake voice (if it hasn`t already) as the largest tributary of the traffic stream running through new fiber-optic pipelines. The Internet serves as one of the biggest engines of this data juggernaut, it is often said. So it would seem to be only a matter of time before the requirements of the Internet, as well as other types of traffic running over Internet protocol (IP), drive fiber into every corner of the network, even to businesses and homes.
Certainly that time appears close at hand to some--but others suggest that the Internet has more hype than horsepower as the driver of fiber into the access network. Whether or not you are a believer appears to depend on a combination of where your product fits into the network, how much faith you put in the development of virtual private networks (VPNs), and how soon you believe true competition will come to the local loop.
The time is at hand
In general, the amount of fiber present in access networks today is relatively small. One could make a case that much of the current new deployment of fiber in this space owes its origin to cable-television companies. Corning Inc. (Corning, NY) says that cable-TV firms ranked as the top consumer of fiber in the United States last year, as multiple service operators (MSOs) and smaller companies expanded their hybrid fiber/coaxial-cable infrastructures in pre paration for providing cable modem and telephony service, mainly to residential customers.
At the same time cable-TV companies extend their Internet- and telephony-driven fiber infrastructures to residences, they may also position themselves to access business facilities, according to Stephen Montgomery, president of fiber-optic market-analysis firm ElectroniCast (San Mateo, CA). Montgomery sees a trend toward businesses moving into suburbs (as well as condominiums and other multiple-dwelling units being built close to office parks). Thus, while the MSOs creep fiber closer to their residential customers, they`ll also find themselves in close proximity to business users. Pointing to AT&T`s recent acquisition of MSO Tele-Communications Inc., Montgomery says it may not be the cable-TV companies themselves that take advantage of this newfound business access.
"Cable-TV companies are going to be in perfect position, in terms of the fiber anyway, toward fiber to the service area, and they`re going to be in perfect position to be acquired by the CLECs [competitive local-exchange carriers]," he predicts. As was the case with AT&T, carriers will find cable-TV companies attractive for both their rights of way and the amount of fiber they`ve installed within the service area, he says.
Of course, carriers such as CLECs, the regional Bell operating companies (RBOCs), and AT&T`s fellow interexchange carriers must first see a need for fiber in the access networks before they find the MSOs` assets valuable. Few think that the current bandwidth demand--Internet or no--warrants immediate fiber build-outs in access networks and local loops.
"Right now, if you go out there, there`s certainly a demand for OC-3s, but usually in big metropolitan areas where it`s going to the very high-profile, large corporate installations--and they`re not that ubiquitous yet," says Mark Cree, marketing vice president for Neo Networks (Minnetonka, MN). His company has developed a massively parallel networking device, the StreamProcessor, that includes seven application-specific integrated circuits which contain more than 1000 software-programmable RISC processors to provide terabit-level data processing at full-duplex native port speeds as high as OC-48. However, the initial release of the product, slated for late this year or early next year, will carry only Gigabit Ethernet ports with an eye toward application in enterprise networks. "We`re probably like any other terabit player. There`s just not enough fiber going to enough large installations--and, actually, medium-sized installations--to justify mass deployment of fiber-based services yet in the IP world," he explains.
But Cree predicts this situation will change rapidly, thanks to growth in what he calls "the Internet edge." In his view, Internet service pro viders (ISPs), whatever their stripe, will look to provide a wider variety of IP-based services--firewalls and other security assets, virtual private networks (VPNs), and a menu of different quality-of-service offerings to meet varied customer requirements--than simple Internet connection. That will lead to a requirement for more processing power at the edge of the network.
"As we talked to consultants and some of the ISPs, we`ve found out that`s where all the services are going to be deployed; that`s where you`re going to generate your revenue if you`re an ISP--where the corporate WAN [wide area network] gets connected to the Internet, not in the core," explains Cree. "So what you need there is a box or a device that has a lot of processing power, that can support a lot of applications. Because the carriers want to go out there and sell services. If all they can do is give you an IP connection, their revenue model gets pretty flat in five years."
Meanwhile, as users get accustomed to Gigabit Ethernet rates within their enterprise networks, they`ll want links to the outside world (including branch offices) that are significantly faster than the DS-3s now in use in most applications. This requirement will increase as users sign up for the new services offered by ISPs--they`ll need a port through which these services can flow to and from their enterprise networks at high speeds. So even though there is not a huge demand for direct fiber connections to businesses now, "that`s going to rapidly change in the next 12 months," Cree believes. "Because as they get used to these gigabit-speed applications in their enterprise, slowing down to DS-3--which may or may not be fiber-based--becomes unacceptable in a lot of cases, especially in company-to-company applications."
Thus while Cree envisions many of his customers will be ISPs that will install his system at their points of presence (PoPs), there will be some instances where the StreamProcessor finds itself in a corporate backbone, where it may connect to the PoP at OC-3 or greater. He views this burgeoning market as similar to that for Fast Ethernet a few years ago. "Fast Ethernet burst onto the scene, and within 12 months it was in everybody`s backbone," he says. "And I think we`ll see the exact same thing with fiber going to large corporations."
Montgomery agrees that access network scenarios have evolved--particularly due to the extension of enterprise networks, especially local area networks (LANs)--to cover wider areas. Gigabit Ethernet deployment will push enterprise networks closer to service-provider networks, he says, changing typical local-loop topologies. "Because of this LAN extension, it`s no longer a conventional telco local-loop network; it`s actually a point-to-point," he says.
Fiber deployment will increase because of this trend--in fact, it already has, Montgomery asserts. "We have actually been viewing this for at least a year now," he explains. "One of the hottest discussions, in terms of our clients--and our clients are the people who make the systems and components, mainly--has been the access network." Montgomery says that this talk has translated into action. "Right now, we`re looking at OC-3, OC-12, and we`re starting to see OC-48 within the loop," he says.
Not so fast
Yet not everyone is convinced that the Internet edge--or any other part of the Internet--will have much of an effect on fiber deployment within the access network in the near future. Mark Steinberg and Aman Kapoor of Ryan Hankin Kent (South San Francisco, CA), for example, view the Internet`s effect on the access network with skepticism. "Part of this has to do with tariffs. If you look at most of the RBOCs, most of the growth is still DS-1 and DS-3. So I don`t foresee that you`ll see OC-3 and OC-12 growth rates overtake that in the next 12 months," offers Steinberg.
Even in the Internet backbone, speeds and capacity aren`t changing as rapidly as some imagine, he says. "Most of the Internet today is still at DS-3, and you`re only starting to see things like the MCI backbone on the Internet--when they finish the upgrade this year, it will be an OC-12 backbone. And Sprint, on a couple of their links for SprintLink, will do OC-48," Steinberg explains. "But most of the Internet is still at DS-3 and OC-3 and OC-12. And that`s in the backbone, as opposed to the edge."
"We`re certainly not seeing extensive use of fiber as the subscriber outlet in these networks," reports Gordon Saussy, director of marketing for Torrent Networking Technologies Corp. (Silver Spring, MD). Torrent makes equipment that aggregates IP and other traffic for transmission on Synchronous Optical Network (SONET) backbones. "Subscribers are aggregating in on fractional T3 service or fractional T1 service, and there`s a great deal of interest and enthusiasm around very large-scale deployment of megabit-rate service over DSL [Digital Subscriber Loop], leveraging the copper that`s already in there. Certainly, it`s not out of the question that there will be subscriber connectivity directly on fiber. It`s a small, small percentage of the total market for Internet connectivity to business customers."
Saussy doesn`t see this situation changing soon. "If you look at the subscriber network for a DSL environment, what they`re thinking of is that at the subscriber outlet--the office, the home, or whatever it is--you`ve got DSL modems on the copper loop, and that copper loop goes 18,000 ft to the CO. And then when you come out of the CO, you`re in the LEC or the CLEC`s fiber network," he says. "So the question is, what`s the timing of fiber replacing copper in the last 18,000 ft? I guess I`m of the opinion that it`s not happening within the next three to five years, outside of some fairly narrow deployments."
Even in the context of describing increased fiber activity in the access network, Montgomery admits there are limits. "When you get to the distribution and feeders, it just depends on the density as far as where the fiber may be going. In the higher-density areas--the high-business metro areas--we`ve seen lots of deployment of fiber over the last few years by the incumbents. So that`s pretty much set, and it`s continuing," he says. "But once you get into that last mile, then we`re finding that`s still copper. And the only real fiber that`s being deployed there is for what we call `new builds.`"
All in all, however, if you take these limited new builds and compare them to the recent past, Cree may not be that far off, says Tom Nolle, president of market-analysis firm CIMI Corp. (Voorhees, NJ). The key is viewing fiber deployment within the context of both past history and realistic expectations. "I think if you say `an explosion in the access network to feed big customers,` I think that`s a reasonable characterization. I think the problem that we tend to have in the marketplace here is that it`s too easy for us to forget the qualifiers of statements like that, like `to feed big customers.` Our research shows that there are probably no more than 45,000 sites in the US that would ever have any credibility to consume fiber-level connectivity. Now if our definition of an explosion is growth from our current level of OC-3 delivery, which is less than 5,000 customers, to 45,000 customers, then, yes, that`s a reasonable characterization. But there are 7.3 million business locations. If we mean an explosion to cover 7.3 million business locations, we`re smoking dope."
Fiber finds its way
Yet even if those who don`t believe the Internet will shortly lead a fiber explosion in the access network, that doesn`t mean that fiber technology won`t be affected by the Web. "In applications where we can`t push an OC-3 all the way to the customer, the next objective is to justify SONET all the way to the edge office," says Nolle. "And if we can justify SONET technology, packet over SONET, or something like that, all the way to the serving office, then what we need to do is take that SONET and use it to support a large number of T1 customers downstream."
Asserts Saussy, "We`re seeing extensive use of fiber--exclusive use of fiber--as the backbone between super PoPs within a service area, whether that`s a university campus, a metropolitan area, or even a region. And that`s where people are running IP over SONET, IP over ATM [Asynchronous Transfer Mode] over SONET--they`re basically shoving the IP packets across fiber-optic backbones into SONET rings."
Nolle sees VPNs as the key to future fiber deployment in networks running IP. "There are two factors at the very high level that are going to drive fiber deployment. One of them is that you can`t justify fiber deployment unless you have more traffic. And so VPNs represent the largest single source of new traffic that`s credible in all of networking."
Saussy agrees that bandwidth demand has to be present before fiber reaches new market areas. He points to the deployment of fiber in long-haul networks as an example. "This is all a matter of economics, right? The fiber build-outs by the large carriers are driven by how fast they can recover that investment in infrastructure. And with this tremendous uptake in data traffic and the promise of all the applications that will be able to quickly leverage this fiber, as opposed to taking a long time, the economics are there to make it worthwhile," he explains. "This is why companies like Level 3, Williams, Qwest, etc., exist at all, this is why they`ve been able to attract so much funding. It`s because the economics show that if we can provide a lot more bandwidth in the backbone, a lot more bandwidth in the regions, that there are applications ready to use it. So I think that it`s a tremendously exciting thing that all of this growth and applications demand on the Internet is getting ready to use a lot of bandwidth, and it`s accelerating the demand for this fiber to be pulled."
Competition represents Nolle`s other fiber driver. "Competition suggests that individual players will build out parallel infrastructures to serve the same forward demand," he says. "So more fiber gets deployed per customer under a competitive environment than under a noncompetitive environment because each competitor has to have a fully geographically capable footprint in order to sell services. And that means deploying what perhaps might be three or four networks of fiber products in tandem with one another, each one of them belonging to a separate company."
If these factors come together, then fiber will eventually reach deeper into access networks--with Internet playing a role in conjunction with other potentially high-bandwidth services. "Internet is only one of several," says Montgomery, discussing fiber drivers. "We think the main driver is digital networks and some of the end-user applications that digital will bring in multimedia voice, video, and data to the end users."
While the timing of fiber`s advance to the access network remains a matter of debate, the consensus appears to be that services such as the Internet, as well as VPNs, could prompt carriers to follow the lead now being set by cable-TV companies. The percentage of businesses that will require direct fiber connections may be relatively small--but they still represent a lucrative market. u