[chat - expanded version of previous article]
tpsite.com
Speed of Light
Only an all-optical network can truly satisfy our lust for greater bandwidth, but getting there will take longer than some might hope.
Never mistake a clear view for a short distance. That's wise advice from our futurist friend Paul Saffo, but it's too often ignored in the euphoria over the latest hot topic in communications: optical networking. Some day, the networks that connect our telephones and our computers may indeed use light from end to end, instead of electricity, to bring us voices and data. That will be the start of a new information age in which communications capacity seems infinite and free. Some day, perhaps, but not soon. For now, the all-optical network is fantasy. We are moving toward it, but the difficulties encountered along the way say a lot about the pace of turning science fiction into science fact.
Optics is the star of the telecommunications equipment market. Although telecommunications capital spending overall is growing at a modest 8% annually, sales of optical equipment are rising at more than twice that rate, driven by the surging demand for Internet service and the rapid expansion of such upstarts as IXC Communications, Level 3 Communications, and Qwest Communications, which hope to use high-speed fiber-optic lines as a competitive advantage over AT&T and other incumbent long-distance carriers.
Adding to the optical enthusiasm is Ciena, the equipment supplier that in a mere six years has turned its early lead in wave division multiplexing (WDM) into $5 billion in shareholder wealth. WDM, which divides a light beam into many signals, is becoming commonplace in long-distance networks and is beginning to move into metropolitan area networks, too. WDM is creating enormous transmission capacity. Later this year, for example, Lucent Technologies expects to begin commercial shipments of a WDM system that enables a single strand of glass to carry as much data in a second as now travels over the whole Internet in that time--about 400Gb.
Shining path
Now for the next step. The technologies that made WDM possible, the erbium-doped fiber amplifier (EDFA) and the in-fiber Bragg grating, are about to be joined by another advance--the commercial introduction of optical switching. This is one of two essential technologies for taking full advantage of the capacity that wave division is creating. (The other is optical wavelength translation, in which a signal is moved from one wavelength to another as it passes through a converter so that the original frequency can be reused.)
For the most part, WDM is used in point-to-point optical transmission; linking WDM lines to the rest of the telecommunications network requires that the optical signals be converted into electronic form, and usually demultiplexed, before being redirected. Switches are beginning to emerge, however, that can work directly with light. The latest of these form the basis for reconfigurable optical add/drop multiplexers (OADMs) that are due soon from two venture-backed companies, Chorum Technologies and Tellium, a Bellcore spinoff. Like their electronic counterparts, OADMs will be used to connect local communications tributaries with the high-speed Synchronous Optical Network (SONET) fiber-access rings that encircle most large American cities. If they work as expected, they will do the job far more cheaply, because they are much simpler than the optoelectronic devices now used to regulate traffic at these intersections. In the case of Tellium (and also Chorum, we expect) the OADM will be remotely programmable. That's an important advance.
Supporting casts
Both Chorum and Tellium seem to be working on a ferroelectric liquid crystal approach to optical subsystems such as OADMs. Chorum has developed a number of technological piece parts and is now in the process of meshing those into subsystems and determining its approach to the market. The company, previously known as Macro Vision Communications, is moving from Boulder, Colo., to Richardson, Tex. On its board are Jon Bayless of Sevin Rosen, the lead investor in Ciena during its startup days, Flip Gianos of InterWest, and Bob Palluck of CenterPoint Venture Partners. Tellium, with venture backing from Accel Partners, Blue Rock Capital, Oak Investment Partners, and Worldview Technology Partners, is building systems-level optical products for the carrier market, including WDM terminals and amplifiers, as well as cross-connects and add/drop multiplexers. The company's approach to cross-connects is actually a hybrid of optics and electronics, although its marketing materials refer to the product as an optical cross-connect.
Light switch
Optical networks will also require a more complex and general-purpose switching system, the optical version of the digital cross-connects (DXCs) that provide flexibility and reliability on today's networks. Besides linking any of several incoming lines to any of several outgoing lines, these cross-connects reroute traffic automatically when a network path has failed. But today's electronic cross-connects are optimized for voice traffic rather than the data transmissions that will soon account for the majority of all telecommunications.
Accordingly, data passes through today's DXCs at relatively low speeds because network traffic is first broken into small elements (multiples of basic 64Kbps voice units), switched, and then recombined. Take the case of OC-48 traffic, the 2.5Gbps data flows that account for the majority of the bitstreams on today's long-distance, point-to-point optical highways. Current cross-connects, such as the Tellabs 532E, are designed to operate on T-1 lines, each of which combines 24 voice channels. To employ the 532E at OC-48 rates in the simplest case, a 2-by-2 cross-connect, both 2.5Gbps data streams would first be decomposed into 160 tributaries apiece, each poking along at a mere 1.5Mbps, so that the functional role of a 2 x 2 OC-48 switching matrix would, in fact, be played by a 320 x 320 T-1 switch.
Too much of a good thing
Indeed, the progress of WDM is magnifying the cross-connect task. For example, in the fourth quarter of this year, Lucent expects to begin commercial shipments of its WaveStar OLS 400G optical networking system that enables a single strand of glass to carry 80 streams of information at a mixture of OC-48 (2.5Gbps) and OC-192 (10Gbps) rates with a total capacity of 400Gbps. The system can handle as many as eight of those fibers, for an aggregate capacity of 3.2 trillion bits a second. While that's impressive, to make full use of the OLS 400G would require cross-connects with a capacity for 8 streams of 80 channels, or a 640-by-640 switching matrix, which is well beyond the abilities of conventional techniques.
Are we there yet?
Optical cross-connects (OXCs) will solve this kind of problem--and because they operate without the costly DXC burden of electronic conversion, decomposition, aggregation, and photonic reconversion they should also be cheaper. Data communications traffic has been growing much faster than the revenues it generates, so the cost savings that would result from better cross-connects seems very attractive to Internet service providers and other high-traffic carriers.
Despite years of pursuit, however, the optical cross-connect has proven to be an elusive goal. For a little historical context, let's recall that although the laser itself was invented 40 years ago this summer, more than two decades passed before the first system trials of lightwave communications. Not until 1983 did a commercial system actually begin operation. Wave division multiplexers began shipping in 1995, eight years after scientists announced the first practical optical amplifier. (In-fiber Bragg reflectors had been under development for longer still.) Nearly a decade ago, Bell Laboratories tried to make true cross-connects using the so-called SEED (Self Electro-optic Effect Device) technology, but it turned out to be much too expensive.
A few companies--such as Akzo Nobel Photonics (a unit of the big Dutch chemicals and pharmaceuticals company), DiCon Fiberoptics, and E-Tek Dynamics--now offer simple, commercial optical switches that can connect one or two incoming fiber lines to two or more outgoing lines. These are really bypass switches, not cross-connects, however, and only Akzo's is solid-state.
More robust devices are on the way. Last year, Lucent introduced a 32-line, wavelength-selectable OXC based on lithium-niobate electro-optic wave-guide technology, but potential customers wanted much greater capacity--more on the order of 256 or even 512 lines--and the technology wouldn't scale. The issue was cost. To early Lucent developers it seemed that OXCs might best be used to link a couple of WDM rings. In fact, however, WDM has exploded in the long-distance portion of the telecommunications network at nodes where hundreds of routes must be interconnected.
While Lucent's OXC might have managed such nodes, each switch lost so much light that too many compensating amplifiers were needed. Greater switching capacity--72 by 72--is available in an OXC from Astart‚ Fiber Networks, which uses a combination of mirrors and lenses to redirect light from port to port. Although WorldCom is testing the product, it is designed for use with multimode fibers, not the single-mode type of fibers that prevail in long-distance telecommunications.
End of the tunnel
Despite the frustrations, development continues. Large interexchange carriers are exploring the OXC area in an effort to understand the system requirements. But smaller companies are helping to lead the way in basic technology. Nine-year-old Lightwave Microsystems is using semiconductor manufacturing techniques to create microscopic optical switches in the form of polymer waveguides on a wafer of silicon. The path of the light can be switched at rates as high as 40GHz. Lucent and a venture-backed San Diego company, Optical Micro-Machines, are exploring micro-electronic mechanical systems, or MEMS. The microscopic mirrors used in this effort are variations of the light-processing technology that Texas Instruments developed for digital projectors. In an experimental Lucent device--a four-port add/drop system--the microelectronic mirrors tilt under electronic control and either reflect an incoming beam back to its source to resume its travels along the network or deflect it to another route. (See "Soul of a Small Machine," Sept. 15, 1997, for our thoughts on MEMS technology.)
Wallet warriors
Investment interest in optical switching is heating up. Numerous business plans for optical switch companies are circulating among our venture capitalist friends, and several such startups are keeping low profiles in the Dallas area, primarily developing switching fabrics dedicated to OC-48. When we dropped in on an optical networking symposium at Stanford University two weeks ago, we counted 42 venture capitalists in the audience. Company names have started to pop up, usually with few product specifics attached. Lightera Networks (formerly Enfusion Systems), founded by Jagdeep Singh, who headed AirSoft before its sale to Shiva, recently collected $9.5 million in seed financing. Optical Networks, which has raised $7.3 million, is a spinout of video products vendor Optivision and is said to be working on technology from Stanford. David R. Huber, the technical founder who left Ciena in April of 1997, has just received a venture funding commitment for a new optical networking equipment company, Nova Telecommunications.
Given all the development effort, commercial all-optical switches are certain to be on the market within a year or two. What's less clear is the adoption rate. We're not looking for the birth of another Ciena here because of what we hear from optical switching's most likely customers. While they're impressed with the emerging technology, they're concerned about the ability to manage it, to fit it into the networks they already have. After all, the chief expense in telecommunications equipment isn't the purchase price; it's the cost of operation, which predominates by a ratio of three-to-one. And that operational expense is still much too high, say such network operators as SBC Communications and Sprint.
Wait signal
The demand for telecommunications capacity is growing so rapidly and has reached such a level that only a new generation of equipment can meet it. Only optics. But not just yet. Eventually, faster switches using optics will be developed and placed in service--where appropriate. Even then, inertia and economics will keep us on copper wire and electronics for a long, long time. The all-optical enthusiasts are right about the direction of the future; it's the timing they've got wrong. |
