Some recent GG post of note from the Telecosm Lounge
My personal portfolio for the last fifteen years has been dominated by Applied Materials, Qualcomm, and Velocity Capital. 5/21/00
I believe that listening to the technology, as Carver Mead says, still is absolutely critical. I've never encountered anywhere a business analysis as profound and as far-reaching and as true as Clayton Christensen's. From the first time I heard it, it absolutely riveted me, in part because it accorded with profound technological insights which Carver had taught me years before. Clayton's analysis oriented me toward looking to inferior technologies which in new combinations yielded dominant and powerful companies. Thus, the model of low and slow that yielded the initial integrated circuit. Bad, low-capacity capacitors; nonconductive, poorly conductive wires; flawed transistors-all of these inferior devices combined on a silicon sliver that can be manufactured ultimately for 80 cents produced the $300 billion business of contemporary electronics.
And then, too, I encountered a man named Will Hicks, who invented the single mode fiber years ago, and thus is a major figure in the fiber optics business. I met him when he read my book Microcosm, and he wrote me a letter, saying, "This is a dramatic and wonderful story you tell here, but the advance of fiber optics over the next decade is going to be a million times greater than the advance of electronics. Its ultimate potential will be hugely more significant." This really arrested me, because I've always been the most cornucopian thinker in the room. Here was a man with long experience, a stream of inventions, who proposed that I was being very modest in my proposals for the future potential of technologies. I believe that Will Hicks's prophecies are virtually becoming true today. Fiber optics really does portend millionfold rises in potential bandwidth distributed around the globe.
Carver had earlier told me about and alerted me to this phenomenon, because he said he didn't think there would ever be an optical computer; he thought that optics were not compatible with computing, that electrons affect other electrons, and thus can make possible fabulous computing devices. Optics were made by God for communications. So what we have today are not two conflicting technologies where electronics and optics somehow compete and one or the other will prevail. Rather, these are two beautifully complementary technologies where fiber optics supplies the communications at the core and the essentially dumb networks, as David Isenberg has always insisted, while the electronics affords the potential of huge intelligence on the fringes of the network. This process continues, the continued migration of new dimensions of intelligence to the edge of the network, and that model still continues.
The invention that I've encountered that really arrested me, and that makes me think Hicks's prophecy may well come true, is by a company that is not in fact a sponsor here today. Lucent has invented AllWave technology, a major innovation in optics, a breakthrough of considerable importance. I went down to Lucent to try to find somebody who was enthusiastic about AllWave and who comprehended its implications, and I could not discover any. This shows the problem that can occur when a huge, powerful, dominant, and innovative company harbors an innovation. We're trying to invent applications for this fiber that didn't compete with their other fiber products, and thus we are shunning AllWave off into the niches of cable TV, which are not big consumers of their other fiber products.
Every textbook on optics I've read-and I've been reading them in great numbers in recent years-has an attenuation chart, which defines the optimal paths for fiber optic signals. The chart is the pass-band essentially for infrared propagation down optical fibers. It always shows at one end around 1300 nanometer wavelengths, the low-attenuation region. Then there is a water spike mountain, a great region where it is assumed that the intrinsic properties of the fiber absolutely prohibit the propagation of infrared signals. Then, at the other end, is a 1500 nanometer region, another low-attenuation region.
The key thing about AllWave is that it essentially flattens the water spike mountain and creates a coherent of broad band for the propagation of signals from 1280 nanometers all the way up to 1625 nanometers. This increases the number of potential wavelengths that can be accommodated at the usual spacing from 160, which is currently the state of the art that Nortel has announced, to about 1,000 at 25 gigahertz spacing and as many as 2,000 wavelengths at 12.5 gigahertz spacing, which is currently the state of the art. There are several companies that can render working systems with the wavelengths separated by 12.5 gigahertz. This means that you can have 1,000 wavelengths in one fiber thread. Then you can put as many as 864 fiber threads in a single cable. That means 864,000 wavelengths in a single fiber cable. You have 864,000, and each of those can carry as much as 10 gigabits a second. That means a total bandwidth of 8.6 petabits per second down one fiber sheath. This strikes me to be about a millionfold rise in total capacity of fiber cable in a little over ten years. That's thousands of times more capacity than the whole world telecom network commanded as few as five years ago. This fits the otherwise hyperbolic projection by which Will Hicks changed my life and thrust me from the Microcosm into the Telecosm.
The 864,000 wavelengths might be deployed in a single fiber cable. 864,000 wavelengths seems to me to signal that we are returning to a system of circuit switching. Everyone speaks of the complete domination of the industry by package switching-sending little packets of data in envelopes with addresses on them. But optics is dumb. Optics can't read addresses. Optics can sort wavelengths. Because optics can't read addresses, if it's going to be an all-optical network, it's going to have to be a dumb network. It's going to have to switch and shift lambdas-wavelengths-and that means a return to circuit switching. Except on the edge of the network where electronics ends again and the intelligence will have to reside, the center of the network will be of circuit switch networks. We will be able to waste bandwidth. When you have 864,000 lambdas in every cable and there are hundreds of cables, these can be fed in conduits all over the country where fiber optics is deployed.
This is the new abundance. I focus on abundances and scarcities as the key governing forces in the evolution of technology. I think that the new abundance is going to be circuits and wavelengths, and you're going to be able to waste these wavelengths. Companies will learn how to shift thousands of wavelengths with passive devices that cost hundreds of times-even thousands of times-less than the great add-drop multiplexors and the terabit routers and all the powerful electronic switching tools that populate the network today. All of those are going to give way to passive optical devices. There will be some active optical devices, and there will be electronic controllers, but essentially what you have is optics creating a system millions of times cheaper and more cost effective than the existing system. It's this kind of technological breakthrough that is the foundation of the huge disruptions which will occur in the communications economy over the next decade.
I'll leave you with that vision of 864,000 lambdas, 8.6 petabits a second. These are huge disruptive forces, because they represent an advance that is far more powerful and more rapid than the continuing evolution of Moore's Law: It reverses the pattern of abundances and scarcities, which until now has governed the evolution of technology. The abundances were transistors, and silicon area, and programmer's time. All of these abundances were at the heart of the economy that has prevailed in recent years. I think what makes possible these waves of disruption is the emergence of new patterns of abundance and scarcity. I think the abundance of silicon area is a factor that has been much underestimated in accounting for the evolution of the computer industry. It has always been possible to plug in another mother board, daughter card, back plane, blade, whatever, in the computer covered with silicon devices. And indeed every five years the silicon area has increased 150 percent. This has been one of the great abundances of the industry in the past. But with the increasing emergence of single-chip systems as the dominant systems in the industry, silicon area has become scarce. Transistors are less abundant. The true abundance which governs the evolution of disruption in the future comes from the vast horizons of bandwidth abundance that are emerging from the fabulous creativity in optics today 5/17/00 from disruption innovation conference
Agilent is obviously not the last word in switching technology and I hope the May issue did not imply that it is. We have enthusiastically described several MEMs switches, led by Xros. But perhaps the TI announcement is a portent; if TI cannot make micromirror switches in volume who can? Agilent supplies a robust, manufacturable technology that seems likely to become widely and cheaply available before MEMs based switches are perfected.
Agilent does not offer doctrinaire 100 percent non blocking, but virtual non blocking in its Clos topology. Thus it escapes the implied combinatorial explosion of a non blocking mesh.
Yes, the bubbles are binary. That's what you want in a switch and in a mesh. In the new lambda based circuit switched networks, 100s of times a second is ample. The inherent add drop functions make the Agilent device useful in many parts of the network.
On the other points, Agilent engineers (and Giovanni Barbarossa, a former Agilent designer of the system, now head of R&D at Avanex) show great confidence in the behavior of their system. We shall see.
The issue of scalability is real, but in my judgement affects all the proposed systems. The MEMs have lover attenuation but pose more exacting manufacturing and maintenance challenges. Attenuation problems can be addressed by the fully tested technologies of amplification.
But the debate is by no means over. 5/13/00
The Avanex Powermux allows wavelength channels to be adapted to the content and bitrate rather than forcing the content and the bitrate to be adapted to the wavelength channels, as in all other optical systems. Thus the Powermux enables slow and reliable mostly passive circuit switching in optics as opposed to multi gigabit per second packet switching by hugely expensive electronic processors. 5/12/00
*******************
The Lucent post seems to have components of an earlier GTR.
Haven't heard the Velocity Capital mention before.
Jack |
