SIMON CAO'S VISION (from Telecosm conference)(rough draft)
The key objective of this very short remark I am going to make this morning is to set the stage for a very controversial debate about what the optical network is, about what the future optical network is. I am going to do that using the simplest terms, and I hope by doing that everybody in the audience can understand what optical networks are, and future optical networks in 8 minutes. I am going to do that with a direct comparison to a story about, a train, a car, and a plane.
In my opinion there a 3 different kind of versions of optical networks, a direct analogy to traditional transportation systems. Let me start with the most traditional ones, the trains network, what I call. The key feature is the train is dumb. The track is controlling all its destinations. That ladies and gentlemen, is our current based telephoning system. Basically you're sending photons without any intelligence, and designing all the intelligence on the control surface, traffic control centers. It requires manual or electro-mechanical switching to change the route, and it must be refueled very frequently, its very difficult to off load contents, but however it does carry a lot of cargos.
The challenge of using the optical train network first of all is the railway....it must be a smart optics.... a hub that is where the opto-electronics big fat router comes in, Basically what happens in this big fat router is what happens in the train station in Jenkentu, Japan. People get off the train, and you see millions of people going to different directions of the trains, and on that platform, the Japanese have the most impressive signs to tell you where the trains are going, where you want to go, and which platform it's going to be, what time it's going to be, and people have to wait there, and go up the escalator and go down the escalator to the direction of the platform.. and that is exactly what happens in those big opto-electronics big fat routers. Imagine, they're millions of routers there, imagine they're are millions of trains coming in there, that would be very challenging to build this kind of big fat routers.
Another story I can tell you. A lot of you probably have been to Washington National Airport. Have you ever experienced when you go down the airplane, you have to take those stupid buses from one terminal to another terminal? That is exactly what happens in those opto-electronics routers, a very fast train comes in, and you have to go into very slow and less modularity buses.....You go all the way up and sit and wait there for 1/2 an hour before the bus goes in...those rickshaws are those buses....
Then let's go to another scenario, stations, where there is smaller stations you need to drop off cargos, passengers, that's currently achieved by SONET add/drop. They do a very good job of that. Then the advantage from the SONET add/drop is during a track control and switching the track so it can route the train to a different direction.... This is where the optical switch goes in, starting from Lucent slamming and sliding mirrors,.... or micromachine mirrors, to Xros, Nortel....to Agilent's champagne bubble switch, that is where they come in. All they're doing is moving the tracks..... and using optical repeaters to refuel the engine, to water cool down the engine,...... Then comes the train, train is very dumb, very easy to make optics dumb, photons its very hard to control them, so its very understandable that people start with this kind of approach, because its very hard to make photons smart. The only thing you have to do is hook up the multi-cargo, and you can do it by electronic TDMs, SWDMs..... and the most difficult challenge is how do you implement the brake, to brake the differential speed control, when the high speed train taking a small radius turn you need to control the speed to the inside wheel to the outside wheel. Normally in the fiber world it's addressed as a dispersion management problem, and the most important of all .......is communication between the train and the traffic center............
I would say the next alternative for the network is a car network. A car can move very fast, sometimes faster than a train, particularly when you are driving a Ferrari like that. It can change lanes very easily, only if you are good drivers, but you require a pre-existing map, navigation system, signage, and signaling to change routes. It is relatively easy to refuel and more frequently, but you cannot carry too much, however you still see millions of trucks going across the nation. The analogy is what a lot of people talk about the future..multi-protocol, lambda switching optical network. Is this a switchless network? It is a switchless network in this case if not something change the road, ... and the photon change some kind of property by itself and therefore change the lane.
That is the fundamental difference between a train network and a car network.
The challenge of the car network is to deal with dumb optics, the road, multi lane freeways.. passive SWDMs... car pooling...., access ramps: ODAMs, freeway junctions: optical cross connects.. gas stations: optical receiving, in fact the same thing we do the same thing as the transport business has been doing about hundreds of years, except..... . The driver is the smart photon, it will require very sophisticated navigation systems, now days in high class cars it happens by punching in where you started and where you want to end up, it will automatically guide you there, basically this function will be achieved by electrical routing. One of the most important things is you must have a steering wheel, the most easy to steer a photon is changing its frequency or changing its wavelength. This will require tunable lasers and more importantly tunable wavelength converters, also you need speed brakes differential speed control, single line of traffic control will be relying on development of the optical technology of ...... require only local level management systems.
And the third version is the airplane version of the network, and some people think its the most ultimate version of the network, where sky's the limit, there is no lane, its much faster, you can go across the ocean without refueling.............. , can move easily in any direction because you don't have limitations of a freeway or a train and you can drop off packets or packages wherever you desire without stopping the train, sometimes we can parachute down to the Ferrari, right down to the multi-lane freeway. so let me resummarize the challenges of the airplane network, and the bigger challenge is to deal with big fat dumb optics, to deal with the sky, unlimited wavelength channels, most of these results we have achieved in our photonic plant in Richardson doing a demo we can successfully demonstrate a 1000 wavelength on a single fiber, and on each wavelength we can optically separate and add/drop another 100 fractional wavelengths giving a total 100,000 frequency channels, ladies and gentlemen, just think about that, sky is the limit, you have to easily do the air drop on the air catch system to do optical add/drop on the fly, the airport hub would again will be the optical cross connect, and the long reach in-air refueling.....that is the secret of the ultra long haul network, and the pilot once again must be even more smarter, navigation system, steering wheel, speed control, and the air traffic control.
The key of the open architectual optical network is dumb optics, and smart photons. Dumb optics means indifference, means insensitive, means nonoptimized, means transparent, means scalable, means open architecture, means universal, means reliable, means always available. Smart photons means intelligence, packets of consolidated information.. router and infrastructure......
At the end, the question you be most interested in, were to bet your money....... well just looking back through history, did the train network get killed? no its still here.. Did the airplane network totally dominate? no.. so 3 networks can co-exist and each network may be particularly suitable for particular applications, for example...one such application is the long submarine length between China and U.S, perfect for train network networks, because you don't worry about breaking all the traffic... so shoot down the bullet trains through the underseas tunnel, and on the other hand in the metro access network, obviously the car network is the optimum solution, because here bandwidth is not the issue, the priority number one issue is connectivity, accessability. The road, the freeway, the streetway is easiest way to connect to your home, and in the national backbone network where its fully meshed and at the same time we have medium size traffic to transport accross the nation, it will ideal for the airplane network, In any case the big winners must be the open architectual network, and in my opinion that means dumb optics, smart photons.
I absolutely agree with George........it means the end of the bits per second, means the end of digital signal processing domination, means beginning of new era of photonic signal processing, and finally the network becomes a computer.
Thank you |
