A Death Foretold
Cisco and others that supply the telecosm will implode...if they don't embrace the all-optical network.
By George Gilder
--------------------------------------------------------------------------------
Last year was the biggest sales year ever for SONET, the synchronous optical network hardware used for routing voice and data packets over the telco's extensive fiber-optic networks.
A.D. 2000 likely will be even better, and the market may grow to $10 billion...right before it collapses. Like the final supernova of a brilliant star, SONET will blow up to thousands of times its original brightness--just before it caves in on itself to a dark, silent death. And like the resulting black hole, SONET can suck companies as large as Cisco Systems and Nortel Networks, Lucent and AT&T, and as fashionable as AMCC and PMC-Sierra, into its downward spiral, unless they commit fully to the new world--the all-optical network.
Until the late 1970s, no one had to "interoperate" telecommunications devices from different vendors whose products had different protocols. You just used machinery provided by Ma Bell and her minions around the globe. Today, by contrast, thousands of companies supply electronic components that move digital packets among phones, faxes, computer modems, and everything else digital. These complex, highly profitable electronics are installed at many transfer and repeater points along the fiber-optic lines that carry the bulk of telephone and Internet traffic.
For years, Cisco's routers could interoperate with more protocols and interfaces than anyone else's, giving the company a clear lead in connecting the many players and their technologies. But a photonics earthquake is shaking up Cisco City and the other electronics suppliers.
The current telecom network architecture is based on the Open Systems Interconnect model. It has seven functional layers and can be thought of as a set of seven envelopes inside envelopes, each of which must be opened in sequence as a data packet passes through the network. SONET, organized in "rings" along the fiber lines, is integral to moving the packets through the sequence.
At the heart of most industry arguments is whether the seven layers are too many or too few. Adding a layer--of middleware, encryption, or access, for example--often rings IPO cash registers and creates lucrative new niches. Removing a layer, or more-- for example, by using new wave-division multiplexing (WDM) technology--can mean amputating dozens or even hundreds of companies that supply those levels.
Better photonics equipment can switch light directly and send signals much farther without repeaters. WDM can vastly multiply the number of signals carried over existing fibers. Put together, these advances will eliminate electronics along the fiber network altogether. Only a modicum of good ol' silicon circuits will be needed at the endpoints.
When that happens, SONET will collapse. Suppliers like Cisco, Nortel (which commands 41% of the SONET market), and Lucent (which, after acquiring Ascend, has an estimated 33% of the more modern SONET market) will have no buyers for this gear.
In certain circles all-optical networks remain highly controversial, since most of the industry's expertise revolves around complex electronics protocols. SONET and other clever electronics are seen as necessary to guarantee "quality of service." However, with several million times more reliability and more potential capacity than electronics, optical networks largely trivialize all the quality of service guarantees that justified those protocols.
So, the real issue for investors in networking companies is just how many of the seven layers will be eliminated by the optical juggernaut. Cisco is central to most of these questions. It currently is moving deeper into the optical domain. However, with Intel's purchase of Softcom Microsystems, Cisco faces a major new challenge from below. Meanwhile, Ciena, Corvis, Chorum Technologies, Sycamore Networks, Xros, Avanex, and Optical Networks are driving SONET to the edges of the network, where it will wither. Even Nortel with its powerful acquisition, Qtera, is now trying to eliminate everything but the optical layer from the center of the network. These companies are marginalizing interoperating issues--together with all the pastel boxes and seven-tiered towers of the old model.
The key to the shining new city on the hill is wave-division multiplexing. WDM is the crucial technology driving the bandwidth blowout. With the arrival of WDM, SONET rings become Nortel nooses (and Lucent lodestones). As Desh Deshpande, chairman of Sycamore, explains, a SONET ring is like a railroad line with no express trains. Not only does every train stop at every station but every passenger must get off at every stop and trundle over to the stationmaster to show a ticket, to get approval either to leave the station or get back on the train. At the next stop it is all repeated again, perhaps 20 or more times coast to coast.
The cost of the SONET boxes that do this work has been as high as $200,000, with each one filling a bay 7 feet high. Though the Cienas and Cerents of the world are shrinking both price and size, the calculus is drastically worsening with the onset of WDM. Each ring typically has 8 to 10 boxes for each pathway. But every time the carrier lights another wavelength in a fiber, on a SONET ring, it must buy and install another 8 to 10 boxes on that ring. With state-of-the-art Nortel gear, which features 160 wavelengths, this means 160 ring elements on each fiber costing hundreds of thousands of dollars. Speeding up the train won't help because every box in the ring has to be upgraded, since the equipment is bit-rate sensitive.
This bottleneck can't endure long. Because SONET has placed a choke-hold on the potential of WDM, even the most conservative voices in the optical industry are ready to strangle SONET in return and collapse all the telco protocols into the all-optical network. Presenting a swan song, even Sprint, previously a fervent advocate of SONET and ATM (asynchronous transfer mode), agrees with the anti-SONET, all-optical consensus.
Hermetically sealed and whisked undisturbed along optical fibers, optical signals seek to avoid the sophisticated processing and protocol shuffling that represent the pride and potentiality of electronic networks. With wavelength routing, perhaps 80% of packets can pass end to end, leaving the electronics to manage only the remaining 20% that must be specially processed. These drop-off bitstreams will be small enough to be handled by realistically scaled electronics, such as a Cisco router or Nortel Edge Switch, serving a single campus, town, skyscraper, AOL server farm, or Global Center data warehouse.
Accelerating the demise of SONET are dispersion management tools and modulation schemes from leading-edge suppliers such as Corvis, which allow optical signals to travel some 3,200 km without being electronically regenerated. The previous need for regeneration every 600 km or so was one of the bulwarks of SONET and electronics at layer two. Since the signal had to be converted to electronics anyway every 600 km--roughly the distance between major cities in the United States as well as in Europe--it made sense to switch packets electronically as well. But if Corvis can fire signals to go coast to coast without regeneration, the major justification for electronic switching goes away.
Eliminating regenerations, however, won't enable large-scale optical switching unless there is a large-scale optical switch or cross-connect available. Corvis will have one ready for commercial use by end of first quarter; Qwest and Williams participated in field trials.
Monterey Networks, now part of Cisco, offers its Wavelength Router with several features, including an optical cross-connect function. But competitors accuse the company of ideological deviations because the device has...an "electronic core." For shame! Not very electronic, responds Monterey's cofounder, Michael Zadikian. Rather, he argues, the electronics "emulate optical processing." Other companies developing optical cross-connects include Xros and Astarte Fiber Networks (the latter working with Texas Instruments).
The promise of WDM to throw off the SONET noose and multiply cheap wavelengths presents historic opportunity for optics suppliers such as Nortel, Lucent, Ciena, and Optical Networks, among others. Once in place, WDM will provide new end-to-end light paths in milliseconds from unused wavelengths. Customers will no longer have to wait half a year to buy or lease a new T1 or T3 line. By 2001 it will be common to buy fractional wavelengths, in real time, for contracts measured in hours or even minutes.
With their massively electronic routers, Cisco and its rivals ultimately face a showdown between Moore's Law, which is doubling the processing power of integrated circuits every 18 months, and the forces of the telecosm, which are boosting optical communications power at least four times as fast. As fiber data rates move from gigabits to terabits and beyond, electronic packet sorting becomes orders of magnitude more complex, while wavelength routing becomes ever more practical and elegant.
This trend, as Cisco CEO and President John Chambers has noticed, pushes the action toward optical networks. He has moved with astonishing speed and determination to reposition Cisco, through its acquisition of Monterey and its staggeringly priced $6.9 billion purchase of two-year-old startup Cerent.
However, the Cerent 454 (now known as Cisco ONS 15454), the company's sole product, is a SONET box. Admittedly it is the most stupendously efficient, versatile, diminutive, and altogether wonderful SONET box ever, at half the price of last year's boxes. Bit rates can also be swiftly and cheaply upgraded. And it incorporates post-SONET capabilities. Not surprisingly, the 454 was highly successful at attracting customers; more than 100 signed up in less than nine months, for a projected annualized run rate of $100 million before its first birthday. Among the eager buyers were Williams Communications, Frontier Technologies, Qwest Communications, and Nextlink.
But the party will be short-lived. Almost anyone could sell SONET technology in 1999. That won't be true in 2001. The money flowing into SONET out of telco profits has made carriers desperate to kill it off. SONET is like a blackmailer, working on high margins as long as it can. But blackmailers also suffer one of the highest violent death rates of all the criminal professions.
George Gilder is a contributing editor of Forbes ASAP. He also publishes the monthly Gilder Technology Report. For a newsletter subscription, call 1-800-888-9896. For more information about GTR, email gtg@gildertech.com.
|
