Toward an all-optical Internet

Nov. 1, 1998
16 min read

Toward an all-optical Internet

Optically switched, optically tuned, optically transmitted, and optically amplified...."That`s the ultimate future at every level of the network," predicts Biswanath Mukherjee, head of the computer sciences program at the University of California at Davis and author of the authoritative engineering text, Optical Communications Networks. But Mukherjee readily concedes that the optical Internet is more conceptual than concrete. "You find aggressive use of fiber and advanced techniques of optical switching and crossconnect at the backbone among long-distance carriers, but as for public data networks operating at the local level--well, I can`t name a single one in North America, though they exist abroad."

The unfamiliarity of an acknowledged expert with actual deployments is indicative of the immaturity of the optical Internet and the many hurdles it still must overcome. However, such deployments do exist. So, in fact, do concerted efforts in the area of standards, though they have not been extensively publicized.

"Almost all of the work in standards for public optical data networks is currently focused on the local loop," relates Paul Schumate, executive director of broadband local access networks for Bellcore. "There has been a lot of activity in this area, and standards for the external aspects of networks are already in place and embodied in Bellcore`s TR-909, which we established in the early `90s. And since then members of the Full Service Access Network Initiative [FSAN] have been working on standards at the interface level and are near to completing them."

The FSAN, consisting of telecommunications carriers such as British Telecom, GTE, Deutsche Telekom, France Telecom, Bell Canada, and others, has proposed comprehensive standards for fiber-to-the-curb, fiber-to-the-cabinet, and fiber-to-the-residence in support of a passive optical network (PON) architecture and Asynchronous Transfer Mode (ATM) protocols. Recently, the organization has invited a handful of manufacturers to join as well; however, standards have not been finalized.

First light

Such focus has also not resulted in any unified approach to extending fiber into the local loop by either equipment vendors or service providers. "You`ve got various proposals such as our own Any Media Ac cess System to push fiber deeper into access," relates Doug Broad, a Lucent spokesperson with expertise in optical networking. "But, unfortunately, there`s not much we can talk about in terms of deployment yet. NTT in Japan is taking the lead in this area, and we`re providing them with equipment, but there`s not a lot of activity in the U.S."

Major infrastructure providers agree. No one is currently ordering a significant amount of optical networking equipment with a view to supporting high-speed Internet or intranet access services, and such deployments that have occurred are largely self-contained efforts involving isolated business parks or exclusive planned communities requiring new construction. Only one competitive local-exchange carrier (CLEC), San Diego-based FirstWorld, is actively building out comprehensive area-wide optical infrastructures and then linking them together optically to construct the kind of 21st century information expressway so many have prophesied. And even the FirstWorld system is largely ad hoc and built on a brokered peering arrangement with NaviSite, a nationwide Internet service provider (ISP). So far, the largely fiber data backbone and the handful of fiber-to-the-curb and fiber-to-the-residence networks are not organically linked in the sense of sharing protocols or defining seamless all-optical data paths, and thus the two network segments do not constitute the basis for a next-generation Internet infrastructure.

And yet two current developments inspire fiber hardware suppliers to hope. The first is the extraordinary proliferation of high-speed access services and the migration of investment capital to such services and the manufacturers who provision them. Second is the growing presence of CLECs in the local loop.

"Access providers are ultimately going to need their own infrastructure," insists Jeffrey Carlson, president of Optical Solutions (Minneapolis, MN), one of the prime movers in promoting fiber in the loop. "When they decide to buy it, we`re ready. They can`t depend on reselling capacity from incumbents indefinitely, and for new construction, fiber is the best proposition. If you assume that more bandwidth will be required, the benefits outweigh the costs, and it`s relatively future-proof."

The local loop in context

Long before the public Internet existed there was fiber at the backbone in abundance. Sprint and MCI, the pioneers in long-distance telephone competition, built fiber infrastructure for themselves and for their resellers through the 1980s. Today, the long-distance giants continue to lay fiber to meet the ever-increasing volume of data traffic, most of it over the public Internet.

Lately these pioneers have been joined by Frontier, Qwest Communications, Level 3 Communications, and others--post-Telecom `96 startups that are busily installing fiber over cheaply purchased rights of way--and by various public utilities that are using their oftentimes extensive networks of dark fiber to provide public voice and data services. Where as recently as a year ago experts feared that the explosion of Internet traffic would soon overload the public switched telephone network, the steady deployment of new fiber has eased congestion on the backbone and forestalled concerns about any imminent collapse of the Net. Furthermore, the rise of premium ISP services tying quality-of-service provisions to pricing has permitted infrastructure owners to recoup investment quickly and has ensured that new fiber will continue to be installed to meet increasing demand for capacity.

Parallel with the stiffening of the backbone has been an upsurge in new network technologies for high-speed access to the backbone. Cable modem access is now attracting a considerable number of subscribers--one quarter of a million to date according to Phoenix-based Kinetics Research, which tracks the cable data phenomenon--and regional Bell operating companies and CLECs throughout the nation are busy rolling out competing digital subscriber line (DSL) services. Less well known (and certainly less well established) but potentially formidable competitors are the millimeter microwave services, including point-to-point and local multipoint distribution services (LMDS). Companies such as Winstar, Teligent, and Advanced Radio are especially active in this arena. Finally, a number of satellite networks have begun to offer high-speed access, including Hughes Direct PC, NEXTAR, GE Spacenet, Digital Xpress, and Westinghouse Wireless Solutions.

High-speed fiber access in this context is indubitably dead last in the number of current subscribers and in the level of carrier activity. Where fiber would appear to be a natural, it is almost a nonstarter. Fiber promoters struggle to explain the discrepancy.

"There`s a considerable amount of POTS [plain old telephone service] over fiber in the local loop," observes Matthew Steinberg, director of optical networking for Ryan, Hankin, Kent, a consultancy headquartered in South San Francisco, "but it has nothing to do with the Internet. Nynex has over 100 thousand lines and US West more than 400 thousand, but it isn`t broadband. True, BellSouth is doing a system with Reltec with terminations at 10Base-T Ethernet connections, but that`s all that immediately comes to mind. Five years ago, there was a lot of talk about fiber-to-the curb, most of it involving video dial tone, but that didn`t pan out. We see a lack of willingness on the part of cable carriers to pay the infrastructure costs, and so, from being the number one prospect for high-speed local access, fiber has fallen to the point where it is seldom even considered today."

William Ecker, a BellSouth spokesperson, confirms Steinberg`s comments. "Basically we`re only doing fiber in the loop for telephony rehabilitation. We`re positioning it for data, but that`s not the application today," he explains.

Staking a claim

So where is fiber actually in place for Internet access? FirstWorld undoubtedly takes the lead, with infrastructure extending over Southern California`s huge Orange County, though not yet blanketing the region. Formerly SpectraNet International, the firm commenced operations in 1996 with a contract to provide complete citywide fiber-to-the-curb to Anaheim. Though that ambitious undertaking has yet to be completed, the firm has since installed piecemeal fiber Internet connections throughout the county.

Jim Schwartz, vice president of product management for FirstWorld, explains the company`s strategy. "We`re basically a LEC, and we have a number of core fiber backbones in the region," he says. We have been concentrating on medium-sized businesses lodged in large structures where we can terminate the fiber in the building`s telecommunications closet and serve multiple customers on the premises."

Initially, according to Norma Thorsen, a former analyst with ElectroniCast, FirstWorld sought a diverse market for its fiber connections. "But it hasn`t happened as fast as we predicted," she says, "and certainly not fiber-to-the-residence."

"We are no longer fiber purists," Schwartz says of his company`s current plans for fiber. "We are very aggressively installing DSL services throughout Los Angeles and Orange County, and we are evaluating LMDS. Frequently we`ll terminate fiber into a DSLAM [digital subscriber line access multiplexer] within a commercial building."

Schwartz is guarded concerning the current market potential of fiber-to-the-curb. "We see mainly business applications such as transparent LAN connections over a WAN with 10-Mbit throughput maintained across the network. High-quality videoconferencing would be another application. We`re also looking at installations in hospitality settings. We`ve put a system in the Anaheim Marriott where the fiber terminates at a DSLAM with very short runs to the exhibit areas. Trade exhibitors are very enthusiastic."

Optical Solutions also reports a number of individual deployments around the nation, though nothing approaching the concentration of FirstWorld`s Southern California Network. "We call what we do `fiber-near-the-home` rather than fiber-to-the curb," says Carlson, "and we see CLECs as our logical customers. The ideal would be to do IP or ATM voice and also to support analog CATV services." Carlson adds, "Most of what we`re doing now are beta sites. A greenfield build to a new community in a remote site without previous infrastructure is the type of scenario where our services become attractive. We recently completed putting fiber in one such community in Rye, Colorado."

Another fairly extensive fiber network providing high-speed Internet access as well as video broadcast is Intercable`s system installed in Alexandria, VA. "You could call it hybrid fiber/coax," says Wayne Davis, vice president of engineering for the Denver-based Intercable, "but I wouldn`t. The usual hybrid fiber cable system puts hundreds of homes on each fiber node and might have amplifiers cascaded 10 deep. We put 150 homes on each node, and we don`t cascade amplifiers at all through most of the network. We`re the closest thing to a true all-fiber cable-television system in the U.S."

Will Intercable extend the approach? "Where the density of the population makes it economical," states Davis. "It`s not feasible everywhere."

GST Telecom (Seattle, WA) is also active in developing small, isolated high-speed fiber access, generally in planned communities. Some communities have taken the lead themselves in installing such systems. Recently, the City of Palo Alto`s public utilities deparment has started to offer fiber to the residence. But for now, fiber at the edge of the network remains a curiosity, and Internet growth has yet to provide much impetus.

Nevertheless, equipment vendors remain confident about the long-term prospects of fiber Internet. "Historically, it`s inevitable," declares Lucent`s Crowe. "If you look at what happens with LECs, they`ve been gradually extending fiber farther and farther out with optical DLCs [digital loop carrier equipment]. Instead of extending copper to the limit of its usefulness, they`ll put electronics farther out and shorten the copper runs to the residence. If that trend continues, which we think it will, more and more copper will be retired and fiber will become more and more the primary access mode."

Fanning the sparks

No one disputes that all-fiber local networks are expensive relative to competing access technologies, and that price has been an inhibitor. Disputes tend to focus on cost versus benefits.

"Fiber is not as expensive as people think," asserts Carlson. "Bellcore studies show a 10% to 14% premium over copper for new construction. But that`s comparing feet of cable, not capacity. When you figure in the capacity, a fiber network is actually the most economical proposition. Not to mention that you`re future-proofing the network."

Stephen Montgomery, president of the ElectroniCast consulting organization, agrees. "Fiber is the winning solution for new builds."

Matthew Steinberg is not so sanguine, however, at least in the short term. "I believe that in the long run fiber will prevail. A single network for delivery of all services makes a lot of sense, and fiber permits that. But," Steinberg cautions, "right now we need applications to drive the extra cost of fiber in the loop. I`m not sure what they are, though the Internet`s what people talk about today."

Interestingly, the University of California`s Mukherjee suggests precisely the application that Steinberg dismisses, video dial tone, another term for video on demand (VOD). "Fiber is the only medium that can deliver the service in a large network, and I think it is a service with wide appeal," Mukherjee observes.

But Tom Nolle, president of CIMI Corp. (Voorhees, NJ), which frequently consults on fiber and other high-speed networking topics, strenuously objects. "I think the demographics are entirely inconsistent with the underlying assumptions of VOD. The studies we`ve done indicate that consumers would have to pay for an average of five movies a week to make the business profitable, and, judging from the subscriber rates for premium channels, that isn`t going to happen."

Steinberg himself expresses further reservations concerning the "killer app" of the mid-`90s. "People won`t pay more than $3 a movie, and they`re getting recent releases through near video on demand anyway. That means the service providers have to store a huge back catalog, which isn`t economically feasible today. If there`s any driver, it`s data, not video dial tone."

Detractors of the VOD concept draw further comfort from the experience of Time Warner`s Full Service Network, which was a huge beta test done in the city of Orlando, FL, from 1994 to 1996 over a vast virgin installation of fiber-to-the-curb. Plagued with technical difficulties and customer indifference, the Full Service Network was never commercialized, and the interface equipment was simply removed at the conclusion of the test.

And yet commercial VOD is starting to appear both in the United States and abroad, generally in conjunction with high-speed Internet access, not in opposition to it. Both Singapore Telecom and Suburban Cable in Pennsylvania have had systems up and running for the better part of a year, and Time-Warner Cable and Adelphia cable have announced massive deployments across their extensive holdings along with cable modems. Still, most advocates of fiber-to-the-curb pin their faith in pure Internet access, not closed-loop entertainment services.

At least one authority, again Tom Nolle, is determined to dash such hopes as well. Says Nolle, "Internet and VPNs [virtual private networks] are the drivers, yes, but they don`t require fiber in the loop or any really high-speed access method for that matter. Our studies indicate that roughly a 200-kbit/sec throughput represents the upper limit for Internet access, beyond which further improvements go unnoticed. Why go any higher? In any case, you can`t, not with fiber, not with any other high-speed access technology, because the cost of T1 connections to the Internet has remained very consistent at $2,200, on the average. When you`re buying access, you`re really buying a fraction of that T1 rate. That`s the limitation, not the last-mile technology."

Still, some observers see a larger phenomenon that could speed the adoption of fiber, and, coincidentally, its fastest competitor, millimeter microwave. Says Suresh Arora, vice president of Hughes Network Systems and head of that company`s LMDS project, "The demand for bandwidth is simply unceasing, and this can be demonstrated by studying usage patterns. I postulate a Moore`s Law of bandwidth [doubling every 18 months], which manufacturers must and will meet. Only two technologies are really in the running, microwave--terrestrial and satellite--and fiber. These will constitute the networks of the future."

Donald Crowe, director of switching and access for Lucent`s access archetecture division concurs. "So long as the supply of bandwidth is not fixed nor is it the monopoly of one carrier, the price of bandwidth has to come down," he theorizes. "That fosters the high-speed applications, and that in turn fosters fiber in the loop."

Perhaps so, but, for the present, fiber-based Internet access with its lack of standards, high installation and equipment costs, and seeming overkill in terms of current user requirements, remains challenged. For fiber, Moore`s law cannot be enacted any too soon. u

Dan Sweeney is a freelance writer based in Burbank, CA.

Fiber and the Internet--evolution of the backbone

As recently as a year ago, the data networking publications were full of dire warnings concerning Internet capacity problems, including predictions that the ever-increasing volume of data traffic would shut down the public switched telephone system and paralyze communications in the United States. Many of these prognostications focused on the situation at the Internet backbone where the National Science Foundation had relinquished control to large telephone companies in 1996 and where coordination was presumed to become subsequently almost nonexistent.

In fact, Internet congestion continues to be a serious problem for both business and recreational users, but solutions are at hand, and the need for an Internet2 with a centrally administered fiber backbone appears to be overstated. The really intractable problems in access speed reside at the local level, exactly where the fiber pavement ends.

The solution to problems at the backbone is not Internet2, but rather what might be termed "overnet." The largest providers of backbone services, such as WorldCom, Sprint, and AT&T, as well as quite a number of so-called premium Internet service providers (ISPs), including Concentric Networks, Frontier GlobaLAN, AboveNet, Digital Island, and NaviSite, no longer adhere to the routing schemes of the public Internet. Instead, they have established private paths for data that avoid the congested public access points, the network access points, and Internet exchange points, in favor of data exchanges at restricted access switches owned by the fiber providers.

Access to these high-speed switches is attained through private or brokered peering arrangements among service providers, sometimes for a fee and sometimes for reciprocal services. Data passed through such switches endures fewer "hops" before reaching its destination and generally keeps to Asynchronous Transfer Mode over fiber signal paths where latency is minimized. In some cases, special routing software like AboveNet`s ASAP will be employed to further expedite data traffic.

The emphasis among such service providers is on improving network management at the hubs rather than increasing overall fiber capacity. Certainly additional capacity will be needed as data capacity grows, but fiber installation at the backbone is proceeding rapidly and currently does not appear to be the principal limitation. What is needed instead is improved routing and switching electronics and protocols beyond simple IP for prioritizing packets and controlling latency.

Premium ISP services are generally not available to the public at large, but are resold to ISPs or offered to business users at premium prices. Because the service providers can guarantee throughput speed, latency, uptime, and data-loss levels, the premium service packages are meeting with increased acceptance.

Jim Freeze, a senior telecom analyst at the Forrester consulting group, has studied this phenomenon extensively and predicts that premium Internet services at previous prices will become ubiquitous. "Businesses are starting to rely on the Web for mission-critical activities, and the best effort mentality of the Internet doesn`t cut it anymore. They`re demanding stringent service-level agreements, and the more aggressive ISPs are ready to give those to them."

Sign up for Lightwave Newsletters
Get the latest news and updates.