The Ultimate Backbone

A survey of ultralong-haul DWDM developments: * Economics * Technologies * Vendors

October 9, 2000

22 Min Read
The Ultimate Backbone

By Scott Clavenna, President of PointEast Research; Director of Research, Light ReadingIt is no coincidence that carriers are beginning to complain they are spending far too much money on network equipment and seeing too little revenue in return. Keeping up with capacity demands on their backbones means deploying high-capacity DWDM (dense wavelength-division multiplexing) systems and the requisite Sonet multiplexers to bring traffic onto the network. While providing bandwidth relief, these deployments entail many consequences, namely massive space and power requirements at core hubs and expensive regeneration along long-distance routes. More importantly, they do not transform the network architecture; they only scale it.

The upshot: While the Internet is fundamentally changing the face of networking, networks are keeping up the old-fashioned way, by throwing expensive bandwidth at the problem.

Simply put, carriers need new optical transport solutions that economically scale in at least four dimensions: channel count, capacity per channel, optical transmission distance, and “real estate” (space and power consumption in a hub). In addition, the ideal transport solution offers a range of choices for each dimension, giving operators the flexibility they need when planning a backbone network. Finally, optical transport solutions must also provide all-optical networking capabilities to allow carriers to break out of the traditional model of network scaling and begin building on advances in optical switching, add-drop multiplexing, and ultralong-haul (ULH) transmission.

It is an immensely tall order to expect a single vendor to address all these requirements equally well, but carriers are finding that at the core of their networks, true network solutions are preferable to a number of niche elements that may or may not be interoperable. The three elements of what we would term the “ultimate backbone system” will include transport, regeneration, and switching.

In the following article Light Reading takes a look at recent developments along the bleeding-edge of ULH transmission and at the future opportunities awaiting ultralong-haul carriers in the “optical services” era.

Why Now?
The Emerging Market
The Outlook
Today, carriers are nowhere near the ultimate. Instead they are deploying 40- and 80-channel DWDM systems that require EDFAs (erbium doped fiber amplifiers) every 80 kilometers and complete electrical regeneration every 400-500km. At nodes where traffic is being dropped from the DWDM line system, vast numbers of Sonet/SDH add-drop multiplexers (ADMs) and digital crossconnect (DCS) ports are required.

Why didn’t carriers push for ultralong-haul sooner? For the last three years, while submarine cable operators have gone all-optical, carriers have continued to deploy standard DWDM systems that required electrical regeneration. There are four reasons for this:

  • They were already saving a great deal of money deploying DWDM, benefiting from its ability to optically amplify many optical channels over a single fiber.

  • The solutions used by submarine cable operators were designed with submarine cables in mind, which had perfectly spaced optical amplifiers, optimized fiber, and ideally controlled environments. In terrestrial networks the exact opposite is true.

  • The network is changing. The preponderance of data traffic on backbone nets is creating much longer transmission circuits, express circuits. These circuits interconnect core network equipment, such as switches, routers, and crossconnects. Voice-centric traffic tends to be highly regionalized, while data traffic (particularly IP traffic) is highly distributed, with many paths passing from coast to coast. Only when carriers found they were being asked by ISPs and other network operators to provision circuits across their networks at distances greater than 1,000km, did it become clear they could benefit from ULH systems.

  • They couldn’t understand the gain until they’d felt the pain.

    The implied benefits of ultralong-haul are obvious. If transmission circuits are tending to be longer than 1,000km, point-and-click provisioning is impossible if 3R (retime, reshape, retransmit) regeneration is occurring along a path. Provisioning a circuit manually takes months, and ISPs are getting impatient. If a network backbone based on ULH is in place, a circuit can theoretically be established (at least physically) in seconds when the network administrator finds the necessary path and restoration capacity in the network through a software tool. Since all regeneration is optical, technicians do not need to manually configure the requisite crossconnects at intermediate nodes. Fewer “chair swivels” means fewer instances for failure, rapid service provisioning, and many more satisfied customers.

    The implications of this fairly straightforward development are enormous. If optical circuits can be provisioned instantaneously across a carrier backbone, then the next era of optical networking is at hand. This “optical services” era will leave behind static provisioning of transport capacity (which is largely what today’s “wavelength services” and capacity leases do) and embrace truly intelligent optical networking. The result will be a further fueling of the virtuous cycle of DWDM deployments: Bandwidth demand begets DWDM deployments, which begets new service creation, which begets bandwidth demand. . .

    This cycle is broken today because DWDM deployments are not creating the opportunity for new service creation and are only relieving congestion along the backbone. This is what’s behind carrier complaints that they are spending too much for too little revenue-bearing service in return. If new ultimate backbone systems are deployed, carriers can begin to treat their backbones as service infrastructures, rather than commodities. Optical VPNs (virtual private networks) might in fact become a reality, leading carriers to compete over value-added services rather than price per circuit mile.

    The economics of ULH begin with a simple proposition: If you eliminate electrical regeneration along a route, you reduce network costs, both in absolute terms of dollars per gigabit per kilometer and in the more relative terms of operations costs. ULH began in the submarine cable market, where submerged regenerators were both costly and less reliable than all-optical counterparts. Undersea systems were the first to employ ULH transmission, though they had the benefit of precisely spaced amplifiers, optimized fiber, and a controlled environment.

    The environment in terrestrial networks is anything but controlled. Amplifier huts are rarely perfectly spaced, fiber type may be different along different routes, and environments are difficult to control. Carrier’s carriers fiber swaps only exacerbate these issues, making the use of envelope-pushing fiber technologies difficult undertakings.

    Quantifying the economic benefits of ULH can only be done on a carrier-by-carrier basis, with information on vendor pricing not readily available today, but there are key items to consider that have driven the proliferation of startups addressing this market.

  • Reduction in 3R regeneration leading to fewer line cards Each OC48 transponder in a typical long-haul DWDM system costs roughly $40,000 today, OC192 transponders nearly $100,000. A 2,500km route carrying 40 channels of OC192 would require at least 480 line cards using traditional DWDM line systems in back-to-back configurations in regeneration nodes. A ULH system could feasibly eliminate 400 of those line cards from the route, depending on the ratio of express traffic at each intermediate node. That’s a savings of $40 million in line cards alone.

  • Reduced operations cost This has always been a difficult value to lead with, but it often helps close a deal. Though operations costs are always difficult to quantify in strict terms, the reduction of electrical equipment from any system tends to reduce its operations cost. Fewer electronic conversions means fewer points of failure, though it may be argued that managing and monitoring a ULH network requires technicians with esoteric expertise. As carriers look to build networks that require fewer “lab coats” per mile, it has yet to be proven that a network based on novel amplification and transmission technologies is in fact cheaper to operate.

  • Space and Power Eliminating 3R regeneration at intermediate hubs in an optical line system will reduce both rack space and power requirements, saving a carrier millions over the life of a cable system. Many DWDM systems house only 16 OC192 lambdas in a single equipment bay. Each one of these bays can consume over 2000 watts of power, which may run as high as $1,500 per month.

  • Cost of upgrade Another difficult one to quantify, but in a ULH network, optical line systems can be expanded by adding cards at end points, rather than all along the route. This speeds capacity expansion and, in the case of wholesale services, time to revenue.

    Since ULH is not just about heroic transmission distances, a variety of technologies are required to create ultimate backbone systems. The primary requirements of such a system are:

  • Distances beyond 500km, preferably up to 4000km, considered the longest span necessary in a carrier backbone This can be accomplished through the use of forward error correction (FEC), Raman amplification, and EDFAs. These technologies are widely accepted among ULH vendors in some combination or another. FEC employs algorithms to greatly reduce bit error rates in long-haul systems, increasing both the possible channel density and span length of a long-haul network. Raman offers the benefit of distributing amplification through the optical fiber plant, rather than at specific locations. This provides much greater control of noise and distortion in a long-haul span, though it puts added stress on the optical fiber and connectors. New gain-flattening filters for EDFAs allow greater numbers of amplifier sites in an optical line system by adjusting for noise accumulation. Some vendors are promoting a hybrid approach, employing both EDFAs and Raman to offer greater reliability.

  • Capacity per channel, or how to get over 10 Gbit/s per channel This is a major challenge, considering signals must also be transported ultralong distances. Dispersion management is the key enabler here, as are new transmission formats such as Return-to-Zero (RZ). Few vendors are addressing exactly how they will support OC768 or speeds above 40 Gbit/s, but a few component vendors are busy addressing the problem (see Vendors Prepare for 40 Gigabit Future. Look for components and modules from CyOptics Inc. (seeCyOptics Targets 40 Gbit/s DWDM Components), JDS Uniphase Inc. (Nasdaq: JDSU), Lucent Technologies Inc. (NYSE: LU), and Nortel Networks Corp. (NYSE/Toronto: NT) in 2001.

  • Channel count and density There are two approaches to increasing channel counts that can be employed separately or in combination. One is closer spacing of channels within the typical EDFA spectral window. DWDM systems today use 100 GHz spacing, resulting most often in 40 channels per fiber. Future systems will employ 50 GHz or 25 GHz spacing using improved filtering technologies and interleavers. Ultimately, systems may move outside the traditional C and extended L bands and employ combined C+L band EDFAs or use a single ultra-wideband EDFA in any one of the EDFA spectral windows. Component vendors will be shipping these products within the next eighteen months.

  • Wavelength stability and accuracy Ultralong transport pushes the physical limits of optical fibers, requiring innovations in gain equalization, dynamic dispersion management, and FEC encoding.

  • Software This is quickly becoming the most important element of any optical networking system. Managing optical networks increases in complexity and places new demands on software platforms once provisioning evolves from a static to a dynamic model.

    Qtera Corp. and Corvis Corp. (Nasdaq: CORV) got the terrestrial aspect of this market noticed in 1999 by announcing complete network solutions around a ULH platform. Qtera’s purchase by Nortel for $3.25 billion (see Nortel Completes Acquisition of Qtera ), followed by the Corvis IPO (see Avici and Corvis Make Stunning Debuts), which resulted in a market cap of $30 billion, created significant hype and fueled the funding rounds of a host of new startups.

    By yearend 2000 at least nine optical networks vendors will have announced solutions for ultralong-haul transport. The group is an interesting mix of startups and stalwarts, leveraging unique technology or adding UTH capabilities to an existing transport product line. Two optical networks names are conspicuously absent: Lucent and Cisco Systems Inc. (Nasdaq: CSCO). Others, like Nortel, have gained this capability through acquisition. Cisco and Lucent will likely make acquisitions in this space in the coming year, as ULH becomes an essential component of any optical network vendor’s product line.

    Segmentation always occurs in a new market, even a niche market. It is happening in the metro space — where vendors are gathering in camps of next-generation Sonet, multiservice DWDM, and metro optical Ethernet — and will happen in the ULH space as well. This nascent market so far exhibits little differentiation, but it will, as the conversations among vendors and carriers about ULH continue.

    How to segment this market today? It’s not simple, but here are a few measures to guide the market watcher:

  • Heroic but inflexible Competitors like to characterize Corvis and Qtera this way, though time will tell if it is indeed justified. These vendors announced they would be delivering systems that supported transmission lengths of 4,000km, Qtera emphasizing OC192 channels, Corvis emphasizing photonic switching along the way. Both have been major financial successes thus far, and Corvis has scored the most contract wins. The problem seems to be in the implementation. Carriers have complained that expensive fusion splices are necessary with Corvis, and its engineers have reportedly acted as though its Corvis’s way or the highway. Those engineers may have good reason. Raman is a tricky technology to work with, and these companies will have the opportunity to improve their technologies as the market develops. May just be sour grapes.

  • The toolkit This approach, favored by Sycamore Networks Inc. (Nasdaq: SCMR) and Ciena Corp. (Nasdaq: CIEN), emphasizes flexibility and interoperability (see Sycamore Goes the Distance (At Last) ). These vendors do not claim to be all-optical like Corvis, (they instead utilize OEO switches at the hubs) but offer the benefits of adding ultralong-reach capabilities only when needed, optimized around different transport distances. Sycamore claims you can’t change the laws of optical physics and there will always be tradeoffs. A mix of FEC, Raman, and EDFA amplification gets the optical signal where it needs to go, from 400km to 4,000km.

  • New tricks OptiMight Communications Inc. has “full spectrum” technology to allow greater power to be used without added distortion by employing a technology that expands the optical spectrum in which a wavelength resides. Others, including Solstis (see Marconi Launches Soliton "Startup") and Algety/Corvis (see Algety Telecom SA and Corvis Boosts IPO With Acquisition), employ soliton or RZ transmission to obtain heroic distances or capacities per channel. Both soliton and RZ methods of transport are based on generating pulses of light that interact with the nonlinearities present in optical fiber to create near-perfect optical waves that can propagate over ultralong distances. These pulses can also be used to allot time slots to optical signals, creating multiple high bandwidth optical channels from a single laser. (For a good history of soliton technology, see Optical innovations never cease, and much of the work that has been conducted for the advancement of science is now being brought out for the advancement of valuations.

  • Hybrid optical/packet solutions This is the newest element of the ultimate backbone market, but emerging rapidly. A number of stealth-mode startups are beginning to appear that bring packet processing right into the optical core. These aren’t terabit routers or optical switches, but unique combinations of the two that create a dynamic optically routed core network (see Village Unveils "Optical Packet Node"). Combined with ULH transport systems, these look like the next iteration of the backbone’s evolution.


    Alcatel SA (NYSE: ALA) is famous for a string of “hero” experiments in optical networking, owing in no small part to its participation in the submarine cable business (see Alcatel Claims DWDM Distance Record and Going Long On Light and Alcatel Claims DWDM Speed Record). At Supercomm 2000, Alcatel reported having set world records for DWDM transmission capacity over ultralong-haul networks. In one trial, the company has achieved transmission of 80 DWDM channels at 10 Gbit/s each over a 3,000km G.652 singlemode fiber, using Raman amplification along Alcatel's 1640 Optinex DWDM transport system. Error-free transmission is achieved through dual-stage hybrid erbium/Raman amplifiers spaced at 80km each, which is typical of most DWDM system installations today. Alcatel has not developed a complete solution for ULH just yet, but will be introducing one in 2001.


    Ciena will be a contender in the ULH market, and is already going a long way towards redefining itself as an optical networks company, rather than a DWDM company. Its DWDM systems are arguably the best available today, and the vendor needs to bring ULH solutions to market to maintain this status. Ciena is incorporating several new software and hardware features — including Raman pre-amplification, FEC, nonlinearity management, dispersion mapping technologies, embedded intelligence, and next-generation amplifiers — to extend CoreStream's range beyond 5,000km without signal regeneration. Ciena expects to field deploy FEC and Raman systems at 10Gbit/s in Q1 2001. The CoreStream ULH system will ultimately be designed to support, for example, up to 160 channels of 10Gbit/s over 2,000km without regeneration.


    Corvis, because of its IPO and first-mover status, is the company against which all the others must position themselves. Corvis has already inked three important contracts, commitments that could total over $550 million. The most certain is Broadwing Communications (NYSE: BRW), followed by Williams Communications Group (NYSE: WCG) and Qwest Communications International Corp. (NYSE:Q) (see Corvis Completes First Field Trials). Corvis was wise to position itself as the provider of complete network solutions, not just ULH. The mysteries and vagaries surrounding its optical switch seem unnecessarily cautious (see Patent Points to Corvis Secrets ), but Corvis is winning contracts and will likely report a strong quarter (see Corvis On Track for First Revenues). The concern: many believe when the rest of these competitors come out of the gate, Corvis may look weak, hampered by its finicky technology and exaggerated claims.


    Nortel got into this market through its acquisition of Qtera, which had been building an ultralong-haul system that promised 4,000km spans that would support OC192 channels in the first release. Nortel has renamed the product OPTera Long Haul 4000 Optical Line System (formerly Qtera Ultra) and is positioning it as complementary to its OPTera Long Haul 1600, which is designed to support 160 channels of OC192. Both are “open” solutions, meaning they can interface with any vendor’s transmission or data equipment. The equipment is being promoted as part of a complete backbone solution when coupled with the OPTera Connect HDX (OEO) or PX (micro-electro-mechanical systems [MEMS] all-optical) switches.

    Nortel is faring well this year with its long haul systems, scoring important contracts in Asia and Australia; in the U.S. with France Telecom, Global Crossing (see Global Crossing Buys Nortel DWDM), Savvis (see Nortel, Savvis Strike $155M Deal), MFN, Williams, and Core Express; and in Europe with BT and Sonera. The Global Crossing Ltd. (Nasdaq: GBLX) contract is the only announced contract for the Qtera product, and is designated for deployment in the fourth quarter of 2000. The west coast portion of Global Crossing's network will initially carry multiple OC192s across a 1,450km route from Seattle to Sacramento, Calif. The optical add-drop multiplexing technology of Nortel Networks' OPTera Long Haul 4000 will allow hand-offs at intermediate sites along the way.


    OptiMight, one of Wu Fu Chen’s companies, is an interesting player in this market space, leveraging a unique technological solution to ultralong haul (see OptiMight Reaches for Distance). OptiMight is interesting. It first appeared on the scene with a solution that seemed targeted at carriers that needed a solution for adding more wavelengths to an existing DWDM network, or adding greater distance to certain spans. Currently, OptiMight is positioned as a complete ULH solution, promoting the benefits of “full spectrum” WDM. This is the brainchild of Ilya Fishman, one of the key research scientists in optical components at Stanford University. Full spectrum is great for marketing: It can be deployed over any fiber, and requires no change in the way a network is installed or managed (a swipe at Corvis). Full spectrum offers the benefits of PMD (polarization mode dispersion) resistance, bidirectional operation, and fiber type independence, and it operates without the need for Raman. This should keep costs low and allow operators greater flexibility in deploying ULH selectively throughout existing networks.


    Sycamore put itself on the map with optical scientists from MIT’s Lincoln Lab (Rick Barry and Eric Swanson), and now a neighboring startup is going the same way. PhotonEx is a stealth-mode company, located (not coincidentally) just a mile from the MIT Lincoln Lab where PhotonEx’s CEO, Kristin Rauschenbach, and CTO, Katie Hall, led a number of renowned optical networking projects. The pioneer architects of BoSSNET, their team pumped 10 Gbit/s over 1,600km across Qwest fiber, and pushed the same rates well over 10,000 km in lab trials. They also developed the world’s first 100-Gbit/s LAN/WAN technology, implementing all optical switching, buffering, and clocking technology. Cofounder Nanying Yin brings a track record of product development fromNortel Networks, where he was Director of Nortel Networks Internet CoreRouter Group.

    Other talent in the company includes RF and high-speed digital experts from organizations such as Boeing High Tech Research Center, Draper Lab, Bell Labs, and Georgia Tech. The product development team is made up of hires from Ciena, Nortel, Cascade, Lucent, Nexabit, Bay Networks, and Ericsson. The PhotonEx product management team includes service provider personnel from WorldCom, AT&T, and BellSouth, among others.

    What is the product? Too early to say, but they are focusing on both ultralong-haul and ultrahigh-speed transport. PhotonEx doesn’t have a “Dan and Desh” to wow Wall Street just yet, but are a long way towards turning the blue sky research of the lab into commercial solutions for the carrier.

    Solinet Systems

    Solinet Systems Inc. is an interesting startup, based in Ottawa, developing an optical transport solution that is meant to act as a second generation to what is currently being offered by existing vendors. Solinet has grown quickly in the past few months and is up to 60 personnel now. Many came from Nortel, predictably, while a recent recruit, Andy Wright of Williams Communications, will bring them serious recognition in the U.S. (see Williams Loses Key Staff ). They are funded by Bessemer, Altamira, and Worldview.

    Solstis (Marconi)

    Solstis is a venture of Marconi Communications PLC (London: MNI), designed to operate as a “virtual startup” from purpose-built premises in Stratford-Upon-Avon, U.K. The Solstis project team will include former members of the Photonics Research Group at Aston University, U.K. The project will utilize dispersion-managed solitons, DWDM, and optical transport technology from Marconi to design ultralong-haul solutions. The company expects the first products to be available in the first half of 2001.

    Sycamore Networks

    Sycamore is growing into a complete optical networking company, and after its recent announcement of the SN 10000, an ultralong-haul system, they now have in place the elements to build optical networks from access to core, with a unified optical switching and routing architecture. Sycamore tends to bring products to market once component innovators have done their job. Sycamore adds design engineering and “soft optics,” the management capabilities necessary to enable optical services. In the case of the SN 10000, design engineering is more prominent than in previous products. Sycamore has chosen the toolkit approach and is offering optional Raman, FEC, and EDFA bands to give carriers a range of options to suit different deployment scenarios. This is meant to counter the prevailing perception that ULH systems are finicky and require complete network redesign around a new platform.


    Xtera Communications Inc. was founded in May 1998 by Mohammed Islam, a professor at the University of Michigan (see Xtera Communications Inc.). Twelve months of operation were devoted to proving the feasibility of Raman-based amplifier technology. In July 1999, Xtera moved from Ann Arbor, Mich., to Sunnyvale, Calif., where the initial prototype was developed and deployed in Alpha trials. Xtera moved its corporate HQ a second time to Dallas to draw on the local talent pool. Jon Bayless is the President and CEO, while other executives are from Fujitsu, Alcatel, and Lucent. Mohammed Islam and Herve Fevrier lead the technology team. Fevrier brings experience from Alcatel’s submarine and terrestrial long-haul optics work.There are no ULH carriers, per se. All will eventually adopt some type of ultimate backbone system; it’s just a matter of time and strategy. Adoption of ULH will be driven by a number of key factors, from improving the economics of carriers’ existing backbone operations to creating wholly new backbones in order to compete more effectively with incumbents. Broadwing, Williams, and Qwest all have significant wholesale operations in place, and it’s no coincidence they were first to embrace ULH.

    Other carriers will undoubtedly follow. Look for announcements from carriers like 360networks Inc. (Nasdaq: TSIX; Toronto: TSX.TO), America's Fiber Network (AFN), Aerie Networks, Level 3 Communications Inc. (Nasdaq: LVLT), Global Crossing, and Core Express. These carriers are in the business of providing capacity to service providers and can greatly benefit from ULH systems. MCI WorldCom and Sprint will become customers in time, to make their backbones more flexible; and AT&T will, eventually — because it has to.

    Ultimate backbone systems are only now being devised, and carriers will move cautiously into this technology because of the critical nature of their backbone networks. The critical elements of ultralong-haul transport — ultrahigh-speed channels and optical switching — remain in their first generation, with many vendors working to push the envelope of optical physics while creating solutions that real-world carriers can deploy and operate without a staff of scientists at every juncture in the network.

    Deployment of ULH will set the stage for all-optical switch deployments. Though we would argue that the optical switch market will be dominated by OEO switches in the coming five years, some level of all-optical switching or add-drop multiplexing will be required in ULH networks to accommodate express traffic at hubs. In combination, ULH and all-optical switching create opportunities for carriers to begin offering a variety of reconfigurable, dynamic wavelength services that more closely reflect the requirements of service providers.

    It is also important to note here that the ULH market is just beginning, and what is commercially available today will soon be dubbed “first-generation” ULH as new solutions are brought to market. In the Internet era, carriers do much more reacting than they do planning. As new switching and routing systems are deployed into backbone data networks, transport networks will be asked to accommodate the needs of these systems. Today, scale and intelligence are required from the transport network; tomorrow’s network will add distance, photonic switching, dynamic dispersion and PMD compensation, and new methods of ultrahigh-speed modulation. Beyond that, it will always come back to flexibility and scale.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like