On the Crest of a Wave
Up until a few years ago, carriers wanting to lease bandwidth from another operator had two options: dark fiber or Sonet/SDH connections. Now a third option – wave services – is getting carriers excited, even in this terrible market, and it’s worth spending some time figuring out why.
The market for wave services was created a few years ago, following the arrival of Wavelength Division Multiplexing (WDM). The technology enables a single strand of fiber to carry multiple wavelengths of light, giving carriers the opportunity to sell some of these extra wavelengths, or lambdas, to other operators, as wave services. The lambdas can be linked together to provide connections spanning metro, regional, national, and international backbones, typically running at OC48 (2.5 Gbit/s) or OC192 (10 Gbit/s) speeds.
Although some of the likely customers for wave services – notably ISPs (Internet service providers), CLECs (competitive local exchange carriers), and Internet data center operators – have hit hard times recently, demand is surprisingly strong. Carriers like Broadwing Communications Inc. (NYSE: BRW), EPIK Communications Inc., Teleglobe (NYSE, Toronto: BCE), and XO Communications Inc. (Nasdaq: XOXO) are reporting brisk sales. And believe it or not, the RBOCs (regional Bell operating companies) are also talking about becoming aggressive buyers of wave services in 2002.
Why? Wave services, it seems, aren’t popular in spite of the market conditions; they’re popular because things are so tough.
Wave services help service providers get a bigger bang from their shrinking capex budgets – not only because they're less expensive than Sonet connections, but also because they can be provisioned much more rapidly. As a result, carriers can respond to customer demands more quickly and avoid buying bandwidth until they really need it – not unlike the just-in-time techniques that help manufacturers reduce inventory costs.
Waves are proving to be a cost-effective alternative to dark fiber or Sonet circuits. Many carriers and service providers continue to expand their networks, both in capacity and in reach. Leasing waves allows carriers to expand into metros, close rings in their backbones, or enter new geographic markets at rates about 30 percent to 40 percent lower than those for leasing Sonet capacity – all without the commitment of purchasing dark fiber and the hassles of bringing that dark fiber to the necessary points of presence and colocation sites within a metro area.
In long-haul networks, a typical unprotected 0C48 lambda costs about half as much as an OC48 Sonet connection, which right now costs about $129 per month per mile in the U.S. In metro networks, a five-mile unprotected OC48 lambda costs about $15,000 a month, while an equivalent Sonet connection costs between $20,000 and $65,000. Even better, lambda prices are currently falling at 20 percent a year, and that’s expected to continue next year as well.
Lambdas, it has to be said, aren’t directly comparable to Sonet, because they don’t offer the same levels of protection and network management. On the plus side, however, lambdas are protocol-independent, enabling users to attach any kind of gear with an optical interface, within the limits of its transponder.
Wave services are more expensive than buying dark fiber, but you get a lot more for your money. Users aren’t faced with having to buy, install, and maintain their own transport equipment. They also get service-level agreements, at least three different protection schemes, and a modicum of monitoring – not up to Sonet’s standards but much better than nothing at all, which is what you get with dark fiber.
Many of the carriers’ carriers I’ve talked with also say they’d much rather sell waves than dark fiber. Why? Because dark fiber, even though plentiful in many routes, is a hard asset, which if sold or swapped is lost forever, often to competition that may do with it what they please. If you sell a wave, you preserve your fiber assets and have greater visibility into how your customers use the service.
The other big strength of wave services is provisioning speed. Wave services can be obtained quickly, often in less than 15 days. That puts them in a totally different ballpark compared to high-bandwidth Sonet services, which can take months to provision. Similarly, getting dark fiber is a slow business, not only because it takes time to organize but also because buying, installing, and commissioning transport equipment takes time.
Wave services also typically have much more flexible contracts, allowing users to take them for relatively short periods of time. In contrast, Sonet contracts are typically for a minimum of a year, and dark fiber lease terms can be much longer.
All of this has an impact on economics, and on service providers’ ability to compete. Wave services make it easier for operators to expand their Layers 2 and 3 networks. They can simply attach routers or switches to the wave and let their bandwidth supplier manage the network. This is where the RBOCs could become aggressive consumers of wave services, as they look to expand their data nets.
The other point to bear in mind about wave services is that the technology is still in its infancy. There’s plenty of scope for further reductions in cost, further speeding up of provisioning times, and further improvements in service options. The technologies likely to play a key role include:
- Optical Switches
Today, wherever waves are aggregated, they require back-to-back terminals to make connections beyond a particular ring. This is very costly and difficult to scale, particularly in metro environments. Using a large-scale optical switch as a hub (both electrical and all-optical core switches can work here – see All-Optical Switching Tutorial, Part 1) allows an operator to manage connectivity among a number of interconnected rings, providing improved economics of service and faster provisioning.
This, in fact, may be the most important new market for optical switches, sitting at the network edge aggregating traffic from metro and regional rings. Much of this traffic could be wholesale lambdas, looking for a handoff to another network operator or another segment of the network.
I would argue here that grooming switches still make sense today, in this scenario, because most of these lambda wholesalers still generate most of their revenues selling DS3s (45 Mbit/s) and OC3s (155 Mbit/s), so they will require aggregation and grooming at the network edge. A system that can perform both grooming and optical switching appears to make the most sense here for the next few years, except in cases where a service provider is focused entirely on wave services and wants to optimize its network around those economics. Then an all-optical switch may prove in, if it is reliable enough and offers a sufficient level of wavelength monitoring.
- Wavelength-Level Monitoring
This really comes in a variety of flavors: from low-cost, small, embedded monitoring at amplifier sites and any location that requires a handoff, to more sophisticated network-level monitoring of wavelength trails through an optical network. Wavelength monitoring, therefore, employs both hardware (optical channel monitors, performance monitors, spectrum analyzers) and control and monitoring software.
This area has received a significant amount of attention this year and fueled the creation of at least 20 startups addressing the monitoring market. Companies like Alcatel Optronics (Nasdaq: ALAO; Paris: CGO.PA) (via its acquisition of Kymata Ltd.), Axsun Technologies Inc., Bookham Technology PLC (Nasdaq: BKHM; London: BHM), Optenia Inc., and Proximion Fiber Optics AB are pursuing embedded monitoring, while Lightchip Inc. is going up the food chain another step, building a whole platform around network-level optical layer monitoring.
Thus far, much has been made of the hardware for monitoring, but a number of companies are beginning to explore the requirements for optical monitoring software. Companies like Emperative Inc., Q-Optics Inc., Syndesis Ltd., and Varros Telecom are looking at managing optical waves system-wide through advanced operational support systems (OSSs).
- Tunable Components
This isn’t necessarily a near-term technology demand for wave services, but it will underpin a new class of wave services that will more resemble optical virtual private networks (VPNs) that provision lambdas between two endpoints. With tunable transmitters, receivers, and filters, network equipment can be built that can dynamically provision, monitor, and switch waves through an optical network. The tunability of these components means a consumer of wavelength services will no longer be purchasing a specific wavelength on a route but purchasing optical capacity at the wavelength level that is switchable through a network. This has been most appealing to storage service providers that may want to provide optical connections to their customers for specific periods of the day, based on their storage access needs.
- Next-Gen Amplifiers
These are particularly useful in metro networks supporting wavelength services. Low-cost, intelligent optical amps that offer automatic gain equalization will allow a lambda wholesaler to offer waves across a network of any distance, so each wave will not have to be engineered or subjected to limitations of existing span design. These amps will come in many different flavors, and can be small enough to be integrated on line cards of metro DWDM systems (Genoa Corp.'s, for example) or as standalone optical amplifier/OADM (optical add/drop multiplexer) sites within a network.
- Optical Signaling
Not a big deal at first, but optical layer signaling (GMPLS, Optical UNI, and the like) will ultimately make it easier to provision wavelengths and manage network resources as they are bought and sold. Optical signaling also makes it easier to offer differing grades of restoration and should speed the delivery of lambdas that span multiple network subsegments through the implementation of Optical UNI and NNI. (For more on signaling, see Light Reading's Optical Signaling Webinar.)
It’s also possible to conceive of a future beyond wavelength services, where the wavelength itself is no longer the commodity, but optical capacity, defined by a user, is the ultimate goal. A challenge lambda service providers will face as these services gain in popularity is “lambda exhaust,” where demand for lambdas quickly outstrips supply, forcing carriers to overbuild on a regular basis. Signs of this are already appearing in Europe, where some of the large pan-European carriers have been selling waves at a brisker pace then their U.S. counterparts.
One vendor, PhotonEx Corp., has an answer to lambda exhaust: Stop selling lambdas and start selling capacity, using GMPLS, LMP, and other emerging standards as tools to intelligently "carve up" core optical bandwidth into consumable increments that can be oversubscribed. This justifies their push to 40G, because it’s easier to carve up bandwidth increments from a fat pipe than from multiple smaller pipes. The true lambda service gives way, in this scenario, to a wholesale capacity service model that scales along a different trajectory, optimizing optical capacity rather than increasing the volume of lambdas sold. PhotonEx is first to admit that this is a few years off, but it’s worth investigating, as the wholesale market rationalizes in the coming two years.
— Scott Clavenna, Director of Research, Light Reading